Category Archives: Work

Personality and Safety: A Review of Beus, Dhanani, & McCord (2015)

Are there certain features of our personality that increase or decrease the likelihood that we will behave safely or be involved in accidents at work? A number of recent meta-analyses have, with increasing rigor, attempted to combine all we know about the relationship between our Big Five personality characteristics (i.e., Openness to experience, Conscientiousness, Extraversion, Agreeableness, and Neuroticism – think OCEAN) and safety at work (Clarke & Robertson, 2005; Chrisitan et al., 2009; Beus et al., 2015). So, what can we say about this relationship?

First, it exists! There is indeed a relationship between four of the Big Five personality characteristics and safety. Surprisingly, agreeableness has consistently shown itself to be the largest and most robust predictor of safe behaviour. I say surprisingly because I would have guessed that conscientiousness, how orderly and responsible we are, would have shown the strongest relationship. Yet, conscientiousness came a modest second place (similar effect size but less variance explained) in terms of predicting safe behaviour. Finally, extraversion and neuroticism showed much smaller but consequential relationships with workplace safety behaviour, such that the more extraverted and neurotic individuals are, the more likely they are to behave unsafely at work.

The idea that agreeableness is the most robust predictor of unsafe behaviour is fascinating and has many implications. Beus and colleagues discuss many of these implications in their paper and I tend to agree with the idea that agreeableness motivates people to seek communion. Essentially, the goal of communion is to seek meaningful and healthy relationships with others. People are thus less likely to behave unsafely to avoid putting others at risk or to make the group suffer because of their own carelessness. In turn, this finding has tremendous implications for the power of social interventions and social features of the organization for shaping safe behaviour. Reframing safety performance as prosocial and adding information about how behaviour affects others to feedback systems are examples of how organizations can improve safety by activating the motivation towards communion.

While I find the idea that agreeableness is the best personality predictor, I could imagine there are almost certainly boundary conditions to this finding. For instance, I could imagine that an agreeable person who finds him or herself in a group that behaves poorly and unsafely will likely be swayed by the group to also behave in a poor and unsafe manner. I cannot think of any empirical evidence off the top of my mind to support this idea, but it would be a fascinating study and one that is entirely possible to do given the advances in multilevel analyses.

Moving on, it came as no surprise that conscientiousness was important to workplace safety behaviour. What was interesting was the idea that extraversion and neuroticism were related to unsafe work behaviour. Beus and colleagues examined particular features of these broad personality characteristics to provide insight into what is driving these relationships. What they found was that the relationship between extraversion and unsafe behaviour was largely driven by a feature of extraversion called sensation seeking; thrill seekers put themselves into more risky situations on average. Meanwhile, the features of neuroticism that were driving the relationship are anger and impulsiveness, while anxiety actually reduced unsafe behaviour. You could imagine the overlap between sensation seeking and impulsiveness to be a particularly dangerous combination of personality when it comes to safety.

While personality and workplace safety behaviour were shown to be are related, what about personality with accidents and injuries? Beus and colleagues address this question both theoretically and empirically. They proposed that personality would be related to accidents and injuries through unsafe behaviour. That is, behaviour would fully mediate or explain the relationship between personality and accidents and injuries. This is what they found*. If you want to understand how personality is related to accidents and injuries, you need to understand how personality drives workplace safety behaviour.

The role of personality on workplace safety behaviour and safety outcomes has become more clear through these reviews. However, there could still be doubt that personality is rendered unimportant when compared to other predictors of workplace safety. For instance, does personality still matter when compared to a robust predictor of workplace safety such safety climate (i.e., the collective perceptions of an organization’s safety practices, policies, and procedures). Again, Beus and colleagues rose to the occasion to provide an answer. In their meta-analysis they showed that personality still mattered when taking safety climate into account (although slightly less than in isolation).

Based on the reviews, especially that completed by Beus and colleagues, we can be confident in a number of things. First, we know what the most important personality characteristics are in predicting workplace safety behaviour and their relative magnitude. Second, we can confident that personality is indirectly related to accidents and injuries through safety behaviour. Finally, we can be sure that personality can explain safety behaviour when taking other predictors, particularly safety climate, into considering.


*NOTE. Well not exactly. Results from Beus et al (2015) suggest that a partially mediated model has a slightly better statistical fit. However, Beus and colleagues suggest that this is only because of how certain personality characteristics are measured, such that some measures are behavioural in nature. Given that the statistical techniques used were not designed to detect and account for methodological contamination, researchers need to make judgement calls as Beus and colleagues did – and to their credit, argued that a fully rather than partially mediated model is appropriate.


Beus, J. M., Dhanani, L. Y., & McCord, M. A. (2015). A meta-analysis of personality and workplace safety: Addressing unanswered questions. Journal of Applied Psychology100(2), 481-498.

Clarke, S., & Robertson, I. (2005). A meta‐analytic review of the Big Five personality factors and accident involvement in occupational and non‐occupational settings. Journal of Occupational and Organizational Psychology78(3), 355-376.

Christian, M. S., Bradley, J. C., Wallace, J. C., & Burke, M. J. (2009). Workplace safety: a meta-analysis of the roles of person and situation factors. Journal of Applied Psychology94(5), 1103-1127.



Criteria for Testing and Evaluating Occupational Safety Interventions: An Overview of Shannon, Robson, & Guastello (1999)

Organizations can sink a lot of resources into improving safety, but how can we make sure that these interventions are effective? Easy – science! Well, not so easy. Especially if you want the science to be done well. Testing safety interventions requires considerable time, effort, and funding. However, if done well, these tests can be invaluable. You may ask, though, what do I mean by ‘science done well’?

Shannon and colleagues (1999) give an inspired answer to this question in their paper by providing a practical but evidence-based list of criteria to consider when testing interventions. I list these criteria below along with some commentary to emphasize the main take-aways.

Criteria for evaluating occupational safety intervention research

 Program objectives and conceptual basis

  • Were the program objectives stated?
  • Was the conceptual basis of the program explained and sound?

These questions are a great place to start. The objectives of the study are key because they will determine which intervention is appropriate and how we would determine whether the intervention was successful or not. Meanwhile, the conceptual basis provides us with a way of organizing our thoughts, making it easier to connect the vast array of moving parts within interventions.

Study design

  • Was an experimental or quasi-experimental design employed instead of a non-experimental design?

As far the hierarchy of rigor goes, experimental randomized controlled trials with baseline measures are at the peak of the pyramid. This is how you would want to design your study to have the most confidence in your results. However this is rarely feasible given the nature of field research. As such, quasi-experimental designs are next down the gradation of the rigor hierarchy.  There are several variations of quasi-experiments that I won’t expand upon here, but there are several great resources to figure out which quasi-experimental features to consider weigh above others (see for example Cook & Campbell, 1979). Finally, non-experimental designs can give you the ability to observe change, but without a control group the ability to infer where the change comes from is hard to determine. These non-experimental designs should therefore remain last case scenarios.

External validity

  • Were the program participants/study population fully described?
  • Was the intervention explicitly described?
  • Were contextual factors described?

Detailed information about the participants (e.g., demographics, means of recruitment, and drop-out rate), intervention (e.g., duration and program content), and context (e.g., current state of safety performance within the given organization) provide crucial insight towards extrapolating any one intervention to other workplaces. As such, it is good practice to be as explicit about these particular features as possible.

Outcome measurement

  • Were all relevant outcomes measured?
  • Were the measurement methods shown to be valid and reliable?
  • Was the outcome measurement standardized by exposure?

The first point brings us back to thinking about the objective of the intervention. In short, your main outcome should be tied directly to your objective. If the goal was to decrease injuries, you should measure injuries. It may also be worth measuring other outcomes, such as those which may help explain how the intervention is working (both the implementation and mechanistic outcomes) and ensuring that it is not having any unintended consequences (reducing major injuries may increase minor injuries or underreporting). Finally, the latter two points ask whether you’re actually measuring what you want to measure (validity), whether your measurement is consistent (reliability), and whether the comparison you’re making is a fair one based on relative exposure to hazards (standardization).

Qualitative data

  • Were qualitative methods used to supplement quantitative data?

By adding some qualitative features, such as interviews, observation, and primary/secondary documents, to the test of the intervention can be very insightful. At a minimum it adds to the richness of the study and at best it can help spot threats to or even strengthen internal validity.

Threats to internal validity

  • Were the major threats to internal validity addressed in the study?

There are a number of potential threats to internal validity (i.e., a condition or event that may lead a research to the wrong answer), with more adding up as the rigor of the research design goes down. For randomized designs, the main threats are those which counteract randomization (e.g., improper randomization), diminish treatment (e.g., unintentional diffusion of treatment to control groups), and reactions of those participating in or administering the study (e.g., going along with or against expectations on purpose). For non-randomized control groups, selection biases are vital to consider. Meanwhile, major threats to the internal validity of before-and-after designs can be those related to the temporal aspect of this design, such as history (i.e., changes attributable to other factors than the intervention), maturation (i.e., natural changes in study group that occur outside the intervention), testing (i.e., essentially the placebo effect), instrumentation (i.e., changes in measurement over the course of the study), and regression to the mean (i.e., individuals in the extremes at one point in the study naturally move towards the average). For more details about these threats, please see Cook & Campbell (1979).

Statistical analysis

  • Were the appropriate statistical analyses conducted?
  • If study results were negative, were statistical power or confidence intervals calculated?

Typically, the analyses for testing the differences between experimental and control groups are fairly straight forward but become slightly more complicated for pre- and post-intervention measurements. Essentially, the less rigorous the study design, the more likely that you will need to add statistical controls to adjust for any differences between groups or within individuals over time.


  • Did conclusions address program objectives?
  • Were the limitations of the study addressed?
  • Were the conclusions supported by the analysis?
  • Was the practical significance of the result discussed?

Finally, the concluding points are good questions to be able to answer after the intervention. By following the criteria above, it will be clear whether the intervention objectives were addressed, what the specific limitations are for the interpretation of the intervention, how this affects the conclusions that can be drawn, and what this means for practice. Striking a good balance will be critical in the conclusion. Field research is still rare, and good field research is even more rare. By following the criteria above and attempting to design a study to maximize the validity of the intervention, the more power it will have towards helping us improve safety in the workplace.


Cook, T. D., Campbell, D. T., & Day, A. (1979). Quasi-experimentation: Design & analysis issues for field settings. Boston, MA: Houghton Mifflin.

Shannon, H. S., Robson, L. S., & Guastello, S. J. (1999). Methodological criteria for evaluating occupational safety intervention research. Safety Science31(2), 161-179.

What We Know About Safety Climate and Workplace Injuries: A Summary of Beus, Payne, Bergman, & Arthur (2010)

What is the relationship between workplace injuries and the perceptions employees have of an organization’s safety policies, procedures, and practices? Beus and colleagues attempted to examine this relationship between injuries and safety climate by meta-analyzing the existing literature. The approach that Beus and colleagues took was particularly interesting because they highlighted a number of important but vastly underappreciated distinctions. They strongly emphasized the theoretical motivation for contrasting directional relationships (i.e., whether injury leads to changes in safety climate, or whether safety climate leads to changes in injuries have theoretical differences) and the separation of climate levels (i.e., we could expect the idiosyncrasies of individual-level psychological climate and the group-level organizational safety climate to have non-overlapping variance to be explained). From a research perspective, the emphasis that Beus and colleagues place on these theoretical and analytical distinctions is strengthened by the fact that they are building on clear limitations of previous reviews on the relationship between injuries and safety climate (Clarke, 2006; Christian et al., 2009).

First, the direction of the safety climate and injury relationship contains two different meanings. In one direction, it is expected that safety climate fosters expectations as to what type of behaviours will lead to certain outcomes. An organization marked by a high safety climate will be one where employees believe that unsafe behaviour is highly frowned upon and that management is committed to ensure that employees maintain a safe work environment. As such, the safety climate-to-injuries relationship will be dictated by expectations fostered by the safety climate. In the other direction, observing or experiencing injuries at work will signal important information with regards to the importance of safety to the organization. As such, the presence or absence of injuries will influence perceptions of safety climate.

Second, Beus and colleagues note how climate can be conceptualized at both the individual- and group-level. At the individual-level we have psychological safety climate, and at the group-level we have organizational safety climate. As Beus and colleagues rightfully point out, very similar phenomena can be quite different at different levels. The idiosyncrasies of individuals, such as their unique experience and worldviews, will distinguish them from the group. However, group norms will shape individual behaviours to make group members more similar in comparison to those in other groups. As such, both individual- and group-level safety climate should have unique relationships with injuries.

Finally, Beus and colleagues note a number of conditional factors on the proposed safety climate and injury relationship. In particular, they suggest that the length of time over which injuries are assessed, the contamination and deficiency of safety climate measures, and the operationalization of injuries will shape the strength of relationship between safety climate and injuries. Length of time over which injuries are assessed was tested as an exploratory moderator, while safety climate contamination (i.e., incorrectly adding modifications) and deficiency (i.e., insufficiently measuring safety climate) should attenuate the climate-injury relationship. Meanwhile, stricter injury operationalization should result in stronger injury-to-safety climate relationships, while broader injury operationalization should produce stronger safety climate-to-injury relationships.

What were the results? Fortunately, there were enough studies to test the injury to psychological (r = -.16) and organizational safety climate (r = -.29) and the organizational safety climate to injury (r = -.24) relationships, but not the psychological safety climate to injury relationship. These results suggest that there is a relationship in both directions, and that the group-level safety climate appears to have a stronger relationship with injuries than the individual-level safety climate.

What about the moderators? Length of time over which injuries were assessed only explained some variance for the organizational safety climate to injury relationship, with longer times of measurement producing smaller effect sizes. Contamination appeared to inflate the prospective relationship between injury to both psychological and organizational safety climates (contrary to expectations), while deficiency weakened these same relationships (as expected). Finally, the operationalization of injuries had a pattern to suggest that stricter operationalization was found to have a stronger relationship in the injury-to-climate relationships, and the opposite for climate-to-injury relationships. However, the confidence intervals overlapped, suggesting that the statistical difference between them is weak.

What is something that managers and organizations can do to reduce injuries and improve safety climate based on the findings of this paper? One finding is particularly important to answering this question. The most common dimension of safety climate is management’s commitment to safety. Beus and colleagues found that this was a much stronger predictor for reducing injuries than vice versa (although the opposite relationship was still significant) and that this was the strongest predictor in reducing injuries than any other safety climate measure. This means that the most important thing that managers and organizations can do is effectively communicate their commitment to employee safety. This can be done in numerous ways, such as providing engaging methods of safety training, safety programs and initiatives, and encouraging the reporting and discussion of safety, accidents, and injuries.  There is little else that can compare to a sincere and invested effort on behalf of organizations to improve the health and safety of employees.


Beus, J. M., Payne, S. C., Bergman, M. E., Arthur, W. (2010). Safety climate and injuries: An examination of theoretical and empirical relationships. Journal of Applied Psychology, 95, 713-727.

Christian, M. S., Bradley, J. C., Wallace, J. C., & Burke, M. J. (2009). Workplace safety: A meta-analysis of the roles of person and situation factors. Journal of Applied Psychology, 94, 1103-1127.

Clark, S. (2006). The relationship between safety climate and safety performance. A meta-analytic review. Journal of Occupational Health Psychology, 11, 315-327.

Consider Dread – A Review of Burke, Salvador, Smith-Crowe, Chan-Serafin, Smith, & Sonesh (2011)

Does safety training work?  It sure does.  But that’s not the interesting question.  Does the type of training matter?  Now things are starting to get a bit more interesting.  The answer to this second question has important implications for both safety as well as the all-mighty dollar.  Are training methods that are highly engaging worth the time and resources they consume?  Or should companies take advantage of economies of scale by letting their employees suffer through poorly made safety videos or e-tutorials?  Previous reviews on this question provide mixed results.  While one paper found evidence that engaging training methods had a much stronger impact in comparison to less engaging methods (Burke et al., 2006), another paper found little evidence for a difference (Robson et al., 2012).

Now what question is typically elicited when faced with mixed results?  It depends – figuratively speaking.  We add to the second question above to make it a bit more interesting: Under what conditions does the type of safety training matter?  This is the question that Burke and colleagues answer effectively in their paper The Dread Factor: How Hazards and Safety Training Influence Learning and Performance (2011).

In this paper, Burke and colleagues readdress the importance of engagement.  However, this time they also provide an interesting boundary condition to the relative importance of engagement.  This is the “realization of the dangers in the work context and associated negative affect” (p. 49) – otherwise known as dread.  The rationale underlying dread as an important feature to consider in safety training has roots in learning through social construction.  A person comes to understand one’s work context and the risks it entails through communication and social interaction with others already engrained in this context.  If there are few hazards in the work context, then the necessity to communicate these risks in such a way as to shape what people perceive to be risks will be less important.  In other words, pre-conceived notions of risk will require less implicit knowledge that can only be shared through interacting with others.

Burke and colleagues maintained that more engaging training methods will still be more important to acquisition of safety knowledge and safety performance.  However, they also proposed that under hazardous work conditions, highly engaged training will show stronger training effects on the outcomes.  Burke and colleagues then set out to conduct a meta-analysis of all the research conducted on this topic to date.

So, what did they find?

First, to cautiously reiterate, safety training matters!  No matter the engagement, safety training has important implications towards the acquisition of safety knowledge and exhibition of safety performance.  Second, highly engaging methods of safety training showed much better outcomes for both the acquisition of safety knowledge, as well as better safety performance.  Third, and most importantly, the level of hazard had a conditional effect on both outcomes.  When hazards were high, engaging methods were by far more important than less engaging methods.  On the other hand, when hazards were low, there was actually no statistically significant difference between the method of training and the outcomes.

So, what are the implications? Investing more into engaging methods of training is well worth it if you operate in hazardous conditions or if the job entails relatively higher levels of hazards to employee safety.  Meanwhile, for less hazardous jobs, it may not be necessary to invest as much resources into more engaging methods of training – at least in predicting safety performance and safety knowledge.  However, keep in mind that there may be other consequences to support the idea of emphasizing more engaging methods of training – even if it is to make something as important as safety a little less boring.


Burke, M. J., Salvador, R. O., Smith-Crowe, K., Chan-Serafin, S., Smith, A., & Sonesh, S. (2011). The dread factor: How hazards and safety training influence learning and performance. Journal of Applied Psychology96(1), 46-70.

Burke, M. J., Sarpy, S. A., Smith-Crowe, K., Chan-Serafin, S., Salvador, R. O., & Islam, G. (2006). Relative effectiveness of worker safety and health training methods. American Journal of Public Health96(2), 315-324.

Robson, L. S., Stephenson, C. M., Schulte, P. A., Amick III, B. C., Irvin, E. L., Eggerth, D. E., … & Peters, R. H. (2012). A systematic review of the effectiveness of occupational health and safety training. Scandinavian Journal of Work, Environment & Health, 193-208.

Re-thinking the Role of Perceptual Acuity – A Review and Commentary on Veazie, Landen, Bender, & Amandus (1994)

This article is a review of the epidemiological literature on occupational injuries spanning 1970 to 1992. While there are brief sections on worker populations (i.e., industries represented in the studies they reviewed), and outcomes (i.e., injuries, ranging from minor to severe), the most interesting and potentially insightful section is on risk factors. The remainder of the article focuses on where research efforts should be directed, such as studying specific risk factors, as well as a thorough consideration of methodological issues related to this line of inquiry.

The worker populations found within the studies Veazie and colleagues reviewed were predominantely from industries that are known to be hazardous or simply accessibile to researchers but not particularly hazardous. Consider manufacturing with regards to the latter.  Manufacturing is an industry commonly examined in research because of the relative practicality of conducting research in this industry and not because of any particular unique hazardous conditions. As such, our understanding of occupational injuries, be it outcomes or risk factors, is potentially industry biased when aggregated.

Meanwhile, Veazie and colleagues contrast the practical motivations with motivations driven by actual hazards, such as the focus on transportation and mining. Both of these industries are recognized as relatively high-risk industries. However, as hinted at above, whether an industry is hazardous does not mean it is more likely to be the target of researchers, as many extremely hazardous industries are underrepresented (at least when this paper was published) as they are less accessible, such as agriculture, logging, construction, and fishing. Veazie and colleagues suggest this is largely due to the “transient and independent nature of their workers” (p. 205). Fair enough.

The next notable section is on the outcomes found in the literature they reviewed. The authors note that most non-injury mishaps are excluded and that most studies focus on accidents. Alas, this causes tremendous grief to other safety researchers as the recording and measurement of accidents does not allow us to separate accidents that result in injuries and those that do not. This inability to isolate injuries from accidents is still a problem for those of us joining the field of occupational safety and searching for empirical precision. Another notable shortcoming of the literature on outcomes that Veazie and colleagues pointed out, and which caused my head to nod incessantly in agreement, was that it has been rare for researchers to isolate severity in their measures of accidents and injuries – an idea that still requires empirical attention.

The most insightful section in this paper is on risk factors. Veazie and colleagues classify three categories of risk factors: human, job content, and environment. Human risk factors include things such as demographics, experience, stress reactions, knowledge, and attitudes (p. 206), job content includes factors such as work design and scheduling, while environmental factors include social and organizational features such as physical stressors and hazards.

Veazie and colleagues outline risk factors found in 32 studies that meet their standard of quality to infer an existing relationship. While these specific factors can be found outlined in the three tables on page 207 onwards, the most insightful comment about human risk factors, and I think potentially overlooked theme for all risk factors, is that these factors in some way influence perceptual acuity. While the authors do not expand on what they mean by this, I think the summary of factors found in one study they mention is worth expanding upon.

Veazie and colleagues use the term perceptual acuity when discussing one study where it was found that noise exposure, hearing loss, and alcohol use (among others) were related to injury in shipyard workers (Moll van Charante & Mulder, 1990). They summarized these and the other factors as those which influence perceptual acuity and left it at that. However, I think this idea could be expanded to explain the connection between not only majority of the human factors, but also the job content and environmental factors.  It can be inferred that human factors infringe upon or dampen perceptual acuity. In turn, this leaves individuals vulnerable to hazards that, under better conditions, they would be fully attentive towards and better able to avoid.

The idea that under better conditions individuals would be better prepared for and able to prevent injuries is only part of the story. Other human factors, such as experience, tell us something else.  For one, less experience is typically related to a higher likelihood of injury at work. This lack of experience plays a role on what people perceive to be hazards in their work place, making them more vulnerable to injury. Perceptual acuity is still an accurate mechanism, but instead of it being hampered attention or focus, it is the nature of the perception. As such, a lack of experience fully well influences perceptual acuity, but not in such a way that under better conditions, fully attentive individuals would be prepared for and able to prevent injuries. Instead, individuals who lack experience would still overlook hazards or engage in work activities that carry job specific risks. How can someone prevent an injury if they are not aware of the hazards or risks or the circumstances that increase the likelihood of increasing these very hazards or risks.

The next sections do not at all discuss the effects that job content nor environmental factors have on perceptual acuity. But again, I argue that these can be roughly described as factors that either infringe upon or shape motivations to allocate attention towards hazards in the workplace. In some sense, this may just me forcibly imposing a way of connecting the factors described in these sections. I argue that job content, represented by aspects of job design and layout (e.g., job difficulty, workload, shift work, and so forth), is largely that which can infringe upon or limit perceptual acuity. Meanwhile I would place environmental factors as those which direct or explain how perceptual acuity is devised and divided, such as physical environmental obstacles, structural incentives, and dealing with other human beings.

While I may be over-simplifying things by stretching the idea of perceptual acuity as something that connects the categories provided by Veazie and colleagues, I found it to be an insightful and fun exercise. Reading through much of the literature on workplace safety has left me thinking there is a lot more room for theoretical improvement. As such, part of my approach in familiarizing myself with the literature has been to take that which we know in different directions – even if I know the odds suggest it will be a dead end.


Moll van Charante, A. W., & Mulder, p. G. (1990). Perceptual acuity and the risk of industrial accidents. American Journal of Epidemiology131(4), 652-663.

Veazie, M. A., Landen, D. D., Bender, T. R., & Amandus, H. E. (1994). Epidemiologic research on the etiology of injuries at work. Annual Review of Public Health15(1), 203-221.

The Past, Present, and Future of Workplace Safety Research – A Review of Beus, McCord, & Zohar (2016)

Beus and colleagues (2016) introduce the integrative safety model to provide a much needed comprehensive and coherent narrative behind research on workplace safety.  This includes thinking which has been supported by research and that which is currently attracting the attention of researchers. The integrative safety model (see figure below) is not an attempt at providing an overarching theory but is simply a way of organizing the most current approaches to workplace safety. The conceptual frame combines the three most dominant theories in workplace safety literature: job performance theory (Campbell, McCloy, Oppler, & Sager, 1993), job demands-resources theory (Bakker & Demerouti, 2007), and organizational climate theory (Zohar, 1980).

Screen Shot 2018-02-24 at 12.39.26 AM.png

An important advance in safety research was to start thinking about safety as a performance behaviour. To perform well, a person requires knowledge, motivation, and skill, and these three are largely determined by individual differences (such as personality) and contextual factors (such as leadership and training). The theory also suggests that history and experience will also shape knowledge, motivation, and skills. Together, job performance theory suggests that individual and contextual factors influence the safety triad of knowledge, motivation, and skill, which in turn influence safety behaviour and outcomes. Outcomes are then said to loop back and have a role in further shaping this safety triad of knowledge, motivation, and skill.

The literature largely, albeit in a rather scattered fashion, supports the proposed links between individual and contextual factors on knowledge, motivation, and skill. These in turn are related to safety-related behaviour and outcomes. However, no actual path or mediation models were reviewed, so it is unclear whether the actual indirect effects have been supported. In addition, the authors repeatedly mention safety skills as a feature with knowledge and motivation, but has this actually been developed? What does safety skill actually look like? The authors acknowledge that the idea of safety skill requires more work, but even the article they cite as an example to have used safety skill (Eklöf & Törner, 2002) really only measured knowledge despite calling their measure knowledge and skill. As such, safety skill is something worth thinking about and potentially developing as a construct, even just to show that it has no effect on safety behaviour.

Another advancement in the workplace safety literature was the adoption of the job demands-resources theory. This theory has proven to be extremely useful because it focuses on an array of job characteristics and contextual factors that either contribute to a person’s ability to do their job (i.e., resources) or contribute to the pressure people face do their job (i.e., demands).  In other words, contextual and job-related demands and resources influence personal resources, which in turn are related to safe and unsafe behaviour.

Again, the literature largely supports this theory. As research on the job demands-resources theory has been fairly substantial, there is a fuller picture of the relationship. Not only do demands and resources rooted in job characteristics and contextual factors indirectly impact safety behaviour through personal resources, they also have a direct relationship with safety behaviour. However, generalizability of the job demands-resources theory is also its weakness. There is very little consensus on how demands and resources interact with each other and what the most important types of demands and resources are to safety behaviour and outcomes. This necessary theoretical contribution will, when it occurs, have important implications for workplace safety research and will have a considerable contribution to practice.

Finally, the most prominent theory in workplace safety research is the application of organizational climate theory in the form of safety climate.  The broader theory suggests that an organization’s collective expectations of how people behave will shape individual- and group-level safety related behaviours.  These expectations typically represent the belief that certain behaviours will be reinforced or punished, and in turn motivate people to behave accordingly. Then, in typical topic specific fashion, the appropriate adjective of safety gets tacked onto climate and we have shared perceptions about the value of safety in the workplace.

The literature on safety climate has turned out to be one of the most productive approaches to explaining and predicting safety-related behaviour.  This includes both levels of safety climate: individual and collective perceptions of safety. However, climate is not the only contextual factor shaping expectations about safety and safety-related behaviour. Other factors include transformational leadership, safety norms, and organizational goal-setting and feedback. While the evidence for these features toward safety-related behaviour is strong, there is disagreement about the intermediate behaviour-outcome expectancy of individuals and the nature of the consequential motivation. The authors argue that safety motivation and behaviour-outcome expectancy produce different types of motivation, the former is a matter of valence (i.e., there is value attached to safety), and the latter is a matter of instrumentality (i.e., the connection between behaviour and outcome is a strategy to achieve or retrieve desired outcomes). Theoretically this makes sense, but empirically I can imagine this would be difficult to separate and is something that will need to be solved to contribute to this argument.

Combining the three theories together, we get the natural tail end of the conceptual model linking individual- and group-level safety-related behaviour to accidents. These accidents in turn have consequences for contextual factors such as policy surrounding workplace safety and perceptions of safety climate. Unlike the previous set of variables, there is no theoretical narrative given to weave these variables together. However, this is arguably unnecessary as it is only one step removed from the previous three theories and can be argued to be a natural consequence of the causal sequence for all three theories.

Nonetheless, the presence of a theoretical explanation for the link between safety-related behaviour and accidents may be warranted.  As much as it is no surprise that safety-related behaviour is related to injuries and accidents, the actual effect size is smaller than would be expected, both at the individual and group level. The authors suggest that part of the story is missing, and other factors outside employee safety-related behaviour play an important role in determining the likelihood of accidents. Therefore, the introduction of a broader narrative encompassing employee safety-related behaviour and accidents will be necessary to fully appreciate the predictors of workplace accidents.

Overall, I found the integrative safety model to be a useful narrative for thinking about workplace safety from a distance. Beus and colleagues provide a good overview of what management and occupational health research has uncovered about workplace safety, what researchers are thinking now, and some speculation as to where we should focus our efforts next. Ultimately, I found this paper to be a helpful exercise to also speculate as to what the future of workplace safety research will look like.


Bakker, A. B., & Demerouti, E. (2007). The job demands-resources model: State of the art. Journal of Managerial Psychology22(3), 309-328.

Beus, J. M., McCord, M. A., & Zohar, D. (2016). Workplace safety: A review and research synthesis. Organizational Psychology Review6(4), 352-381.

Campbell, J. P., McCloy, R. A., Oppler, S. H., & Sager, C. E. (1993). A theory of performance. Personnel Selection in Organizations3570, 35-70.

Eklöf, M. & Törner, M. (2002). Perception and control of occupational injury risks in fishery–a pilot study. Work & Stress16(1), 58-69.

Zohar, D. (1980). Safety climate in industrial organizations: theoretical and applied implications. Journal of Applied Psychology65(1), 96-102.

Chronic motivational state interacts with task reward structure in dynamic decision-making

This paper is about motivation. Cooper and colleagues (2015) claim that the definition of motivation (i.e., “a simple increase in effortful cognitive processing”) is due for a revision.  The authors suggest that motivation is instead better thought of as something more dynamic – an interacting multilevel variable if you will.  This is exemplified in the theoretical lens that they adopted.

The theoretical lens through which Cooper & co. approached motivation is called regulatory fit. This regulatory fit is “achieved when the individual’s global motivational state (chronic or situational) aligns with the local motivational task framing” (p. 41).  When there is “fit”, there should be an increase in effortful cognitive processing and a decreased reliance on habitual cognitive processing. When there is a misfit, the opposite occurs.

To clarify, the global motivational states that Cooper & co. are speaking of are called promotion-focused (i.e., these individuals are more sensitive to potential gains) and prevention-focused (i.e., these individuals are more sensitive to potential losses).

Without overcomplicating things, people who a chronically promotion-focused will engage in effortful cognitive processing if a task is framed as promotion-focused (i.e., they are asked to maximize gains), while individuals who are chronically prevention-focused will engage in effortful cognitive processing if a task is framed as prevention-focused (they are asked to minimize losses).  They call this effortful cognitive processing goal-directed or the model-based system.  Meanwhile, if there is a misfit (e.g., a chronically promotion-focused person is asked to complete a prevention-focused task), people will opt towards the less costly habitual reward-based or model-free system of cognitive processing.

To test this motivational regulatory fit model, the authors recruited participants who were either chronically promotion or prevention focused to repeatedly (250 times) choose between two rewarding options for extracting a valuable resource: one will always provide larger immediate reward but decrease future rewards (called the decreasing option) and the other will always provide lower immediate reward but causes future rewards to increase (called the increasing option).  Meanwhile, participants were randomly assigned to either a gain-maximization condition (the extraction procedures produce varying gains of the resource that needs to be maximized) or loss-minimization condition (the extraction procedures produce a varying output of a dangerous by-product that needs to be minimized). See figure below for how this was shown to participants (gain-maximization on the left, loss-minimization on the right).

Screen Shot 2017-10-29 at 15.11.26

What were the most important results?  In the gain-maximization condition, promotion-focused folks performed better than the prevention-focused folks, and in the loss-minimization condition, prevention-focused folks performed better than the promotion-focused folks.  Even within the regulatory focus groups, the alignment of regulatory focus proved beneficial.  Promotion-focused folk performed better in the gain-maximization condition and prevention-focused folk performed better (albeit non-significantly) in the loss-minimization condition.  The regulatory fit hypothesis of motivation was thus supported.  Additional regression analyses reinforced these findings by showing that that relatively promotion-focused folk performed better in gain-maximization and worse at loss-minimization.

Screen Shot 2017-10-29 at 15.20.13.png


Cooper, J. A. Worthy, D. A. & Maddox, W. T. (2015). Chronic motivational state interacts with task reward structure in dynamic decision-making. Cognitive Psychology, 83, 40-53.

Hand or foot?


Photo via Dr. Joe Kiff

If you could only keep one, which would you choose: Hand or foot? Eyesight or hearing? Arm or leg? Choices like this luckily come to most of us in the form of morbid games of imagination we play with our friends. But for an unfortunate population, the choice is made for them at work.

In an article by Elsie Cheung and colleagues (2003), they drew on an observation of many clinicians: employees who experience severe injuries or amputations to their upper-extremity (i.e., fingers, hands, arms) at work seem to be particularly vulnerable to psychological maladjustment. While anecdotes may serve their purpose, Cheung and co. wanted to test whether those who experienced upper-extremity injuries were in fact psychologically worse-off than others who experienced severe injuries and amputations elsewhere. This clearly had implications for treatment and rehabilitation.

Diving into the library at the Workers Compensation Board of British Columbia, Cheung and colleagues pulled out files for individuals who 1) experienced upper extremity amputations or lower extremity amputations, 2) who were assessed by a clinical psychologist at the outpatient rehabilitation center, and 3) were psychologically healthy prior to the injury.

Statistical comparisons of the two groups revealed some interesting results in line with the observations of clinicians.  Workers who had injuries to their upper extremities had substantially more symptoms of posttraumatic stress disorder (e.g., distressing flashbacks, emotional numbness) and slightly elevated signs of depression. When considering pain, however, both groups experienced similar levels.

Screen Shot 2017-10-17 at 07.40.25.png

So, what is the take away? Why do severe injuries and amputations to our fingers, hands, and arms leave us more vulnerable to psychological maladjustment? Cheung and co. align with Grunert and colleagues (1988), who made the argument that it comes down to functional loss, self-image, and social acceptance. So much of what we do on a day-to-day basis depends on using our hands (like typing this very sentence). What we do is important in shaping who we are, and who we are is who people have come to accept. All of this comes crashing down when that choice is made for the unfortunate few.


Cheung, E., Alvaro, R., & Colotla, V. A. (2003). Psychological distress in workers with traumatic upper or lower limb amputations following industrial injuries. Rehabilitation Psychology, 48(2), 109-112.  

Grunert, B. K., Smith, C. J., Devine, C. A., Fehring, B. A., Matloub, H. S., Sanger, J. R., & Yousif, N. J. (1988). Early psychological aspects of severe hand injury. Journal of Hand Surgery, 13B, 177–180.

Get [M]oving in Mplus – part 5: Define subcommands

Despite having to import a dataset into Mplus from another stats program, you can conduct most of the variable manipulation you need in Mplus. This is good news as you’ll often find yourself in a position of having to transform exisiting variables (e.g., log transformations) or creating new variables (e.g., mean scores).

In any case, it can be very annoying having to go back to SPSS to do all of this stuff. Fret not, Mplus has your back with the DEFINE command.

There are a few notes to make before summarizing the most used operations under the DEFINE command.

  • Operations with the DEFINE command can be done on all observations or a selection of some based on conditional statements (e.g., IF(gender EQ 1) THEN…)
  • Transformations do not alter the original data (phew) but hold the alterations in memory only during analysis (unless you use the SAVEDATA command, then the transformed values are saved)
  • All statements in the DEFINE command are done in order (so if you create a mean score and want to transform it, it must be done in this order and not the opposite)
  • Any new variables you create for use in analysis must be listed after original variables being used in analysis within the USEVARIABLES subcommand.
  • The following logical operators, arithmetic operators, and functions can be used in the DEFINE command:

Screen Shot 2017-06-05 at 6.24.34 PM.png

And here are some of the common operations (although not an exhaustive list) you’ll likely find yourself using at one point or another:

Create mean score variables:

Love = Mean(intimate passion commit); 


Love = intimate+passion+commit/3;

Create summative score variables:

Love = Sum(intimate passion commit);

Create other variables (e.g., interaction terms or convert units such as kilos to pound)

Lust = intimate*passion;
Pounds = .454*kgs;

Grand- or group-mean center a variable or variables:




Standardize a variable or variables:


Transform variables:

Lovelog = log10(Love);
Lovesqrt = sqrt(Love);

Conditional statements:

IF (sex EQ 0 AND relstat EQ 1) THEN group = 1;
IF (sex EQ 0 AND relstat EQ 2) THEN group = 2;
IF (sex EQ 1 AND relstat EQ 1) THEN group = 3;
IF (sex EQ 1 AND relstat EQ 2) THEN group = 4;

If there are other operations that you need to do and are possible in the DEFINE command but I haven’t covered here, please let me know. If there are other operations I ever use along the way, I’ll be sure to update this post!

Rudi[M]entary Model Commands in Mplus – part 3: BY

The third rudimentary model command in Mplus is BY or factor. Although statistically more complicated than the previous two, a factor simply generates a latent or unobserved variable through its prediction of observed variables. In other words, you are telling Mplus you have a variable that exists but cannot be measured directly (what is called a latent variable) and that you have some measurements of behaviour proposed to be caused by this latent variable (what are called observed or measured variables).

This is important to understand, so how about an example?

Consider the personality trait extraversion. People who are extraverted are considered outgoing and gregarious (McCrae & Costa, 1987). However, we cannot put someone’s extraversion on a bathroom scale and weigh it — nor can we pour it out of people into a test tube. Extraversion is simply a way of organizing and thinking about a common pattern of behaviours. In other words, extraversion is a latent variable and we must measure it by gathering observed variables representative of our idea of what an extraverted person is, how they behave, and the thoughts they commonly have.

In psychology, asking people questions about themselves and their behaviour is the most common form of measurement. It is no surprise that people tend to understand themselves better than anyone else (especially when it comes internal behaviours such as beliefs, attitudes, and emotions). When measuring extraversion, we can, for instance, ask people to rate the degree to which they consider themselves as talkative.

There are also other ways of gathering observed variables aside from self-report. We can hire coders to observe someone’s behaviour (e.g., code how frequently a participant approaches strangers to strike up a conversation), recruit people who know our participant (e.g., have peers rate how gregarious our participant is in general), and so forth into the realms of creativity.

Essentially, our model of reality is that the personality trait of extraversion (our latent variable) is causing specific patterns of behaviour, such as talkativeness, sociability, and gregariousness (the observed variables).

Visually, this is what it looks like:

factor example

And here is a generic syntax that would run this factor analysis:

Screen Shot 2017-04-29 at 2.26.51 PM

Now, lets look at an example from a real dataset.

Here, participants were asked to think about themselves and rate the extent to which they agree with the following statements about their tendency to perspective-take (i.e., try to understand the world from another’s point of view):

  1. “I try to look at everybody’s side of a diagreement before I make a decision”
  2. “I sometimes try to understand my friends better by imagining how things look from their perspective”
  3. “I believe there are two sides to every question, and try to look at them both”
  4. “When I’m upset at someone, I usually try to put myself in his/her shoes for a while”
  5. “Before criticizing somebody, I try to imagine how I would feel if I were in their place”

Translating these items into Mplus and producing their factor results in the following syntax:

Simple Confirmatory Factor Analysis;

File is PT5.dat;

Names are PT1 PT2 PT3 PT4 PT5;
Missing are all(-999);
Usevariables = PT1 PT2 PT3 PT4 PT5;

PT by PT1 PT2 PT3 PT4 PT5;
!Latent factor by observed factors

Standardized sampstat Modindices(all);

And produces the following output:

Mplus VERSION 7.4 (Mac)
05/28/2017  10:01 AM


Simple Confirmatory Factor Analysis;

File is PT5.dat;

Names are PT1 PT2 PT3 PT4 PT5;
Missing are all(-999);
Usevariables = PT1 PT2 PT3 PT4 PT5;

PT by PT1 PT2 PT3 PT4 PT5;
!Latent factor by observed factors

Standardized sampstat Modindices(all);

Data set contains cases with missing on all variables.
These cases were not included in the analysis.
Number of cases with missing on all variables:  8

Simple Confirmatory Factor Analysis;


Number of groups                                                 1
Number of observations                                         982

Number of dependent variables                                    5
Number of independent variables                                  0
Number of continuous latent variables                            1

Observed dependent variables

PT1         PT2         PT3         PT4         PT5

Continuous latent variables

Estimator                                                       ML
Information matrix                                        OBSERVED
Maximum number of iterations                                  1000
Convergence criterion                                    0.500D-04
Maximum number of steepest descent iterations                   20
Maximum number of iterations for H1                           2000
Convergence criterion for H1                             0.100D-03

Input data file(s)

Input data format  FREE


Number of missing data patterns             2


Minimum covariance coverage value   0.100


Covariance Coverage
PT1           PT2           PT3           PT4           PT5
________      ________      ________      ________      ________
PT1            1.000
PT2            0.998         0.998
PT3            1.000         0.998         1.000
PT4            1.000         0.998         1.000         1.000
PT5            1.000         0.998         1.000         1.000         1.000



PT1           PT2           PT3           PT4           PT5
________      ________      ________      ________      ________
1         3.990         3.910         4.071         3.684         3.784

PT1           PT2           PT3           PT4           PT5
________      ________      ________      ________      ________
PT1            0.829
PT2            0.559         0.881
PT3            0.496         0.520         0.734
PT4            0.539         0.633         0.481         1.051
PT5            0.574         0.607         0.512         0.671         0.996

PT1           PT2           PT3           PT4           PT5
________      ________      ________      ________      ________
PT1            1.000
PT2            0.654         1.000
PT3            0.635         0.646         1.000
PT4            0.577         0.658         0.547         1.000
PT5            0.632         0.648         0.599         0.656         1.000




Variable/         Mean/     Skewness/   Minimum/ % with                Percentiles
Sample Size      Variance    Kurtosis    Maximum  Min/Max      20%/60%    40%/80%    Median

PT1                   3.990      -0.935       1.000    1.43%       3.000      4.000      4.000
982.000       0.829       0.761       5.000   30.35%       4.000      5.000
PT2                   3.910      -0.797       1.000    1.63%       3.000      4.000      4.000
980.000       0.882       0.368       5.000   28.16%       4.000      5.000
PT3                   4.071      -0.953       1.000    1.22%       3.000      4.000      4.000
982.000       0.734       1.095       5.000   33.20%       4.000      5.000
PT4                   3.684      -0.746       1.000    4.07%       3.000      4.000      4.000
982.000       1.051       0.173       5.000   20.57%       4.000      5.000
PT5                   3.784      -0.664       1.000    2.44%       3.000      4.000      4.000
982.000       0.996       0.001       5.000   25.46%       4.000      5.000



Number of Free Parameters                       15


H0 Value                       -5357.476
H1 Value                       -5337.842

Information Criteria

Akaike (AIC)                   10744.952
Bayesian (BIC)                 10818.296
Sample-Size Adjusted BIC       10770.656
(n* = (n + 2) / 24)

Chi-Square Test of Model Fit

Value                             39.268
Degrees of Freedom                     5
P-Value                           0.0000

RMSEA (Root Mean Square Error Of Approximation)

Estimate                           0.084
90 Percent C.I.                    0.060  0.109
Probability RMSEA <= .05           0.010


CFI                                0.987
TLI                                0.974

Chi-Square Test of Model Fit for the Baseline Model

Value                           2686.639
Degrees of Freedom                    10
P-Value                           0.0000

SRMR (Standardized Root Mean Square Residual)

Value                              0.017


Estimate       S.E.  Est./S.E.    P-Value

PT       BY
PT1                1.000      0.000    999.000    999.000
PT2                1.091      0.039     27.699      0.000
PT3                0.910      0.036     25.272      0.000
PT4                1.099      0.044     24.953      0.000
PT5                1.115      0.042     26.451      0.000

PT1                3.990      0.029    137.334      0.000
PT2                3.909      0.030    130.471      0.000
PT3                4.071      0.027    148.892      0.000
PT4                3.684      0.033    112.616      0.000
PT5                3.784      0.032    118.810      0.000

PT                 0.515      0.036     14.217      0.000

Residual Variances
PT1                0.314      0.018     17.719      0.000
PT2                0.268      0.017     15.958      0.000
PT3                0.308      0.017     18.414      0.000
PT4                0.429      0.024     18.207      0.000
PT5                0.357      0.021     17.217      0.000


STDYX Standardization

Estimate       S.E.  Est./S.E.    P-Value

PT       BY
PT1                0.788      0.015     54.078      0.000
PT2                0.834      0.013     66.497      0.000
PT3                0.762      0.016     48.446      0.000
PT4                0.770      0.015     49.858      0.000
PT5                0.801      0.014     57.054      0.000

PT1                4.383      0.104     42.175      0.000
PT2                4.165      0.099     41.945      0.000
PT3                4.751      0.112     42.475      0.000
PT4                3.594      0.087     41.239      0.000
PT5                3.791      0.091     41.522      0.000

PT                 1.000      0.000    999.000    999.000

Residual Variances
PT1                0.379      0.023     16.487      0.000
PT2                0.304      0.021     14.555      0.000
PT3                0.419      0.024     17.462      0.000
PT4                0.408      0.024     17.170      0.000
PT5                0.358      0.023     15.907      0.000

STDY Standardization

Estimate       S.E.  Est./S.E.    P-Value

PT       BY
PT1                0.788      0.015     54.078      0.000
PT2                0.834      0.013     66.497      0.000
PT3                0.762      0.016     48.446      0.000
PT4                0.770      0.015     49.858      0.000
PT5                0.801      0.014     57.054      0.000

PT1                4.383      0.104     42.175      0.000
PT2                4.165      0.099     41.945      0.000
PT3                4.751      0.112     42.475      0.000
PT4                3.594      0.087     41.239      0.000
PT5                3.791      0.091     41.522      0.000

PT                 1.000      0.000    999.000    999.000

Residual Variances
PT1                0.379      0.023     16.487      0.000
PT2                0.304      0.021     14.555      0.000
PT3                0.419      0.024     17.462      0.000
PT4                0.408      0.024     17.170      0.000
PT5                0.358      0.023     15.907      0.000

STD Standardization

Estimate       S.E.  Est./S.E.    P-Value

PT       BY
PT1                0.718      0.025     28.434      0.000
PT2                0.783      0.025     30.927      0.000
PT3                0.653      0.024     27.098      0.000
PT4                0.789      0.029     27.454      0.000
PT5                0.800      0.027     29.100      0.000

PT1                3.990      0.029    137.334      0.000
PT2                3.909      0.030    130.471      0.000
PT3                4.071      0.027    148.892      0.000
PT4                3.684      0.033    112.616      0.000
PT5                3.784      0.032    118.810      0.000

PT                 1.000      0.000    999.000    999.000

Residual Variances
PT1                0.314      0.018     17.719      0.000
PT2                0.268      0.017     15.958      0.000
PT3                0.308      0.017     18.414      0.000
PT4                0.429      0.024     18.207      0.000
PT5                0.357      0.021     17.217      0.000


Observed                                        Two-Tailed
Variable        Estimate       S.E.  Est./S.E.    P-Value

PT1                0.621      0.023     27.039      0.000
PT2                0.696      0.021     33.249      0.000
PT3                0.581      0.024     24.223      0.000
PT4                0.592      0.024     24.929      0.000
PT5                0.642      0.023     28.527      0.000


Condition Number for the Information Matrix              0.150E-01
(ratio of smallest to largest eigenvalue)


Minimum M.I. value for printing the modification index    10.000

M.I.     E.P.C.  Std E.P.C.  StdYX E.P.C.

ON Statements

PT1      ON PT3                   13.131     0.156      0.156        0.147
PT1      ON PT4                   10.045    -0.117     -0.117       -0.131
PT3      ON PT1                   13.131     0.153      0.153        0.162
PT3      ON PT4                   14.998    -0.136     -0.136       -0.163
PT4      ON PT1                   10.045    -0.159     -0.159       -0.141
PT4      ON PT3                   14.998    -0.189     -0.189       -0.158
PT4      ON PT5                   19.784     0.215      0.215        0.209
PT5      ON PT4                   19.784     0.178      0.178        0.183

WITH Statements

PT3      WITH PT1                 13.131     0.048      0.048        0.154
PT4      WITH PT1                 10.045    -0.050     -0.050       -0.136
PT4      WITH PT3                 14.998    -0.058     -0.058       -0.160
PT5      WITH PT4                 19.784     0.077      0.077        0.196


Use View Diagram under the Diagram menu in the Mplus Editor to view the diagram.
If running Mplus from the Mplus Diagrammer, the diagram opens automatically.

Diagram output
/Users/Granger/Google Drive/Website/Stats Resources/Mplus/Files for post/Rudimentary analyses in

Beginning Time:  10:01:14
Ending Time:  10:01:14
Elapsed Time:  00:00:00

3463 Stoner Ave.
Los Angeles, CA  90066

Tel: (310) 391-9971
Fax: (310) 391-8971

Copyright (c) 1998-2015 Muthen & Muthen


There are two highlighted regions in the output that we want to pay particular attention to. The first region pertains to the Model Fit of our perspective-taking scale (i.e., how well our scale captures reality). Most researchers report the following fit indices: Chi-square test of model fit, CFI, RMSEA, and SRMR. What these mean is a whole other post, but here are the general “rules of thumb” (Hu & Bentler, 1999):

  • Chi-square test of model fit: non-significant (or as small a value as possible — this fit index is unfortunately vulnerable to larger sample sizes, so people can often shrug off a signficant value with the right reference, e.g., Bentler, 1990)
  • Comparative Fit Index (CFI): Equal to or greater than .95
  • Root Mean Square Error of Approximation (RMSEA): Equal to or less than .06
  • Standardized Root Mean Square Residual (SRMR): Equal to or less than .08

In the sample output, you can see that some fit indices meet or surpass our rules of thumb (including the CFI and SRMR) and some fit indices are edging on problematic (including the chi-square test of model fit and RMSEA). Messiness like this is very common in research but the general take-away here is that the scale is satisfactory but not great.

The second region we need to pay attention to is the Standardized Model Results, STDYX Standardization.  Here we have what are called our factor loadings (or lambdas; under the Estimate column) which are kind of like correlations between the observed variables and the latent variable. In general, you want factor loadings no lower than .40, but higher is even better. In this example, our items are loading on the latent factor very well – which is a good sign!

Finally, if you happen to use Mplus Diagrammer instead of Mplus editor, Mplus will produce sweet diagrams such as this to help you visualize your factor analysis:

PT factor diagram

And that is about it for the basics of how to use and interpret the BY command! And now for some Mplus syntax humor: Good BY see you later;


Bentler, P. M. (1990). Comparative fit indexes in structural models. Psychological Bulletin, 107(2), 238-246.

Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1-55.

McCrae, R. R., & Costa, P. T. (1987). Validation of the five-factor model of personality across instruments and observers. Journal of Personality and Social Psychology, 52(1), 81-90.