Multivariate Analyses of Dr. Peter Bentler

Dr. Peter Bentler of UCLA has investigated drug addiction, personality theory, and a host of other behavioral and clinical paradigms. He is also one of the most cited scholars in Psychology for his innovations with Structural Equation Modeling (SEM) and the related quantitative concepts.  In working with SEM, Dr. Bentler  and colleagues created and facilitated a set of procedures for one of the most powerful statistical methods in modern Psychology.

For those newer to Psychology and its methods, it is worth a few moments of praise to elucidate statistics and the importance of SEM. One of the difficulties in conducting psychological research is the control (or lack thereof) exerted by the experimenter on her variables.  In psychological research an experimenter likely controls some things in the study (e.g., the questions that are asked in a social survey, or the presentation of a light source in a visual psychophysical test), but some potentially important variables are uncontrolled  and must be left to vary. One can assume that the participants coming into the experiment will have different genetics and different socialization experiences. The reading level of participants may vary as will their ages, genders, and beliefs. All kinds of experimental variables are simply are not amenable to control in a typical Psychology experiment. If one is unable to physically control the variables, then the best alternative is the use of mathematical analyses that correct for that lack of control.  That’s statistics: the mathematical procedure to determine the likelihood of a particular outcome in an experiment given all that varies with limited or no control.

The specific advantage of using SEM for analysis is not only to account for the variables involved but to use the analysis to determine the relative importance of each of those variables relative to each other. For behavioral scientists attempting to isolate the major contributing factors involved in big issues (e.g., to be able to account for the relative influence of age vs. sex on teen drug use, or personality vs. parenting on suicidal ideation, etc.) is challenging but critically important! And that’s why Dr. Bentler is so widely influential today: he helped to create and promote the tools to make our science better by showing the likely contributing value of the variables involved in our research — even when those variables determined from a tangled web of likely influences that occur within the tested sample.

Dr. Bentler introduces himself:

As noted in his introduction, Dr. Bentler has been at UCLA a long time. In fact, he grew up in Los Angeles and even completed his undergraduate degree at UCLA, maintaining a strong connection to this place and institution for almost the entirety of his adult life. His childhood, however, was divided between early years Germany and his eventual movement to the United States in 1948. Dr. Bentler was born in Berlin in 1938 during the antecedent period leading up to WWII, and the child of parents who needed to seek some escape for his family.

In this excerpt, Dr. Bentler recalls his family’s experience and departure from Germany to England and then to Los Angeles:

As the war loomed, Peter’s family decided it was time to leave Germany. Sadly, the time of open borders had ended, and his family was no longer able to leave once it became clear that it would be safer to do so. They stayed in Berlin for some time and Peter’s father was jailed as a Jew. The family moved to Hamburg for the final years of the war before finally exiting Germany in 1947 for England. After less than a year in England, they set out to the United States and became Angelino’s with a new name and new lives in southern California.

In this next excerpt, Dr. Bentler recalls some of his childhood interests and activities growing up in LA:

There were struggles. Peter and his brother were placed in advanced coursework for which Peter may have been prepared academically, but not socially. He was emotionally and physically quite young relative to his new peer group. So, while he was capable of the academic work, in those adolescent years Peter did not find himself strongly drawn to his studies or particular academic topics. That love of math and psychology which became hallmark of his influence emerged much later. As experienced by many of that era, Peter and his family also had some difficulty adjusting to the fluctuating economy in the States and had to figure out new businesses, new ventures to make ends meet during his childhood.

Dr. Bentler recalls in this next excerpt how he started to find his academic interests and his journey from college to grad school:

Dr. Bentler is not the first to laud his experience at community college on this Journey and he discusses that here. His experience at Santa Monica community college helped Peter to develop his interests and to promote his capacities as a scholar. He completed two years at Santa Monica, then a degree in UCLA. From there, he took on a research position — not as a long term career, but with a bit of serendipity — as an available opportunity for recent graduates with a background in the social sciences. In that position he met someone who suggested that he might do well to seek a graduate degree at the University of Pittsburgh and off he went. Sight unseen Peter forged ahead to seek  this advanced degree, at a new institution, and in a new place. It was a bold move but it ended up being a great one!

Peter completed a Master’s degree in Social Psychology at the University of Pittsburgh with an emphasis on cognitive dissonance. A few things happened during that time in Pittsburgh that led to the next chapter of his life. Peter successfully defended his masters thesis, a project on cognitive dissonance with an appropriate set of experimental checks and controls. Despite the use of a manipulation check and consequently the establishment that participants were experiencing cognitive conflict, the findings were different than originally predicted. In fact, the data challenged the application of the theory of cognitive dissonance in this particular context. One member of Peter’s committee found this all very difficult to accept, and despite approving the methods and paradigm, he thought Peter’s data to be invalid. Inconceivable!

All this made for a frustrating twist for a student to endure: what does it mean if you believe that your research is sound and yet people don’t accept the results? While coming to terms with this professional frustration, things were also changing in his personal life.  Peter had traveled for three months in Europe before starting his studies at Pitt. During that time, he met a lovely women in Berlin and became enraptured. After a courtship and many letters between him stateside and her in Germany, they married and she came to live with Peter in Pittsburgh. Between his experiences and possible frustrations with Social Psychology, and now with ideas of a family, Peter determined that he might switch into clinical Psychology and a new graduate program.

Dr. Bentler recalls moving into clinical Psychology and starting at Stanford University in pursuit of this degree. His memories include mention of several notable figures in Psychology including influences from Dr. Albert Bandura and Dr. Ernest Hilgard:

Dr. Bentler completed his PhD at Stanford in 1964 in Clinical Psychology. His training emphasized behavioral methods such as advocated by Dr. Bandura, personality theory and its inventories, and hypnosis from Dr. Hilgard in support of clinical intervention.  He also had an unique opportunity to supplement his studies with a summer visit at the Oregon Research Institute with Drs. Lou Goldberg and Jerry Wiggins.  He describes that experience below:

This was a brief summer experience, but it immersed Peter in some of the innovative quantitative methods and models being developed in theories of personality. Importantly through this experience and soon after in a post-doctoral experience working with Educational Testing Services in New Jersey Peter’s interest in quantitative approaches to Psychology were nurtured.  After his PhD from Stanford, he spent a year on the east coast, and then quickly applied to UCLA to join their clinical program. UCLA was a chance to come back to the city he knew and importantly, to support his wife. She is a film editor and there is no more active scene for such work than the LA film industry. Dr. Bentler was hired to support the clinical program and his teaching/research were directly oriented to issues in psychological health, understanding, and of direct clinical relevance.

Dr. Bentler started as a faculty member at UCLA in the mid-1960’s where he has remained for more than 50 years!  In this next excerpt he recalls how he got more deeply involved in the central work of his career with psychometrics and statistics:

The problem set matters a lot here. As Dr. Bentler started to weigh into the important issues he hoped to address in his growing set of studies on personality theory and later (and for over 30 years) in drug addiction, he saw the need to improve the analysis tools to understand this data. To understand drug addiction required understanding the skew of its affliction in the population. There are spikes of sub-populations at greater risk for addiction, and viewing this data not as the standard set of normal distributions but embracing those skews and what they meant for predictive variables was critical.  Built from the necessity of creating a better approach to the data for understanding — in addition to the support he was getting from great collaborators and technological advancements in computing — was the context in which Dr. Bentler and colleagues developed the tools for SEM and a related software package to do SEM that they termed EQS.

He describes that mathematical journey to developing SEM and the related stat package EQS in this next excerpt:

Students in psychology programs learn to run regression equations and to make predictions from one variable to another. As Dr. Bentler describes, the beauty of SEM was not just its power for improving prediction but for its accessibility: those involved in Psychology can apply their knowledge of rudimentary regression to an understanding of SEM. In doing so, the predictive value of the data is revealed that much more clearly. Dr. Bentler and colleagues used the challenges they faced in better understanding data to create a package for others, in a similar situation, to benefit from their model.

There is much more to the decades of research and psychometric theory development in Dr. Bentler’s career. I look forward to sharing more with you in the book!

Here are some of the wonderful publications from Dr. Peter Bentler:

Bentler, P. M. (1968). Heterosexual behavior assessment—I. Males. Behaviour Research and Therapy, 6(1), 21-25.

Bentler, P. M. (1968). Heterosexual behavior assessment—II. Females. Behaviour Research and Therapy, 6(1), 27-30.

Bentler, P. M. (1969). Semantic space is (approximately) bipolar. The Journal of Psychology, 71(1), 33-40.

Bentler, P. M. (1972). A lower-bound method for the dimension-free measurement of internal consistency. Social Science Research, 1(4), 343-357.

Bentler, P. M. (1980). Multivariate analysis with latent variables: Causal modeling. Annual Review of Psychology, 31(1), 419-456.

Bentler, P. M. (1983). Some contributions to efficient statistics in structural models: Specification and estimation of moment structures. Psychometrika, 48(4), 493-517.

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

Bentler, P. M. (1992). On the fit of models to covariances and methodology to the Bulletin. Psychological Bulletin, 112(3), 400-404.

Bentler, P. M. (2007). On tests and indices for evaluating structural models. Personality and Individual differences, 42(5), 825-829.

Bentler, P. M. (2009). Alpha, dimension-free, and model-based internal consistency reliability. Psychometrika, 74(1), 137.

Bentler, P. M. (2017). Specificity-enhanced reliability coefficients. Psychological Methods, 22(3), 527-540.

Bentler, P. M., & Bonett, D. G. (1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin, 88(3), 588-606.

Bentler, P. M., & Chou, C. P. (1987). Practical issues in structural modeling. Sociological Methods & Research, 16(1), 78-117.

Bentler, P. M., & Huang, W. (2014). On components, latent variables, PLS and simple methods: Reactions to Rigdon’s rethinking of PLS. Long Range Planning, 47(3), 138-145.

Bentler, P. M., & Newcomb, M. D. (1978). Longitudinal study of marital success and failure. Journal of Consulting and Clinical Psychology, 46(5), 1053-1070.

Bentler, P. M., & Speckart, G. (1979). Models of attitude–behavior relations. Psychological Review, 86(5), 452-464.

Bentler, P. M., & Speckart, G. (1981). Attitudes” cause” behaviors: A structural equation analysis. Journal of Personality and Social Psychology, 40(2), 226-238.

Bentler, P. M., & Weeks, D. G. (1980). Linear structural equations with latent variables. Psychometrika, 45(3), 289-308.

Bentler, P. M., & Wu, E. (2005). EQS 6.1. Encino, CA: Multivariate Software.

Bentler, P. M., & Yuan, K. H. (1999). Structural equation modeling with small samples: Test statistics. Multivariate Behavioral Research, 34(2), 181-197.

Castro, F. G., Stein, J. A., & Bentler, P. M. (2009). Ethnic pride, traditional family values, and acculturation in early cigarette and alcohol use among Latino adolescents. The Journal of Primary Prevention, 30(3-4), 265-292.

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. (*Cited over 50,000 times making it one of the most influential publications among the influential psychologists I’ve spoken with on this Journey!)

Huba, G. J., & Bentler, P. M. (1980). The role of peer and adult models for drug taking at different stages in adolescence. Journal of Youth and Adolescence, 9(5), 449-465.

Ibabe, I., & Bentler, P. M. (2016). The contribution of family relationships to child-to-parent violence. Journal of Family Violence, 31(2), 259-269.

Jennrich, R. I., & Bentler, P. M. (2011).  Exploratory bi-factor analysis.  Psychometrika, 76(4), 537-549.

Newcomb, M. D., & Bentler, P. M. (1988). Consequences of adolescent drug use: Impact on the lives of young adults. Thousand Oaks, CA, US: Sage Publications, Inc.

Newcomb, M. D., & Bentler, P. M. (1989). Substance use and abuse among children and teenagers. American Psychologist, 44(2), 242-248.

Ullman, J. B., & Bentler, P. M. (2013). Structural equation modeling. In J. A. Schinka, W. F. Velicer, & I. B. Weiner (Eds.), Handbook of psychology: Research methods in psychology (pp. 661-690). Hoboken, NJ, US: John Wiley & Sons Inc.