"Failures to detect moderating effects with ordinary least squares-moderated multiple regression: Some reasons and a remedy": Correction to Morris et al.

Reports an error in "Failures to detect moderating effects with ordinary least squares-moderated multiple regression: Some reasons and a remedy" by James H. Morris, J. Daniel Sherman and Edward R. Mansfield (Psychological Bulletin, 1986[Mar], Vol 99[2], 282-288). Several errors went uncorrected. On page 283, the second line of the first full paragraph should read "in Equation 3...." On page 284, in the eighth line of the first full paragraph, the power in the equation should be "1/2," not "12." On page 287, in Table 4, the heading for column 6 should read "Adjusted SS for deletion of X?X?," not just "X?." The heading for column 7 should read "H?: β?=0c, partial F," not "β?." Finally, in line 3 of the table note, "X?X?" should read "X?,X?." (The following abstract of the original article appeared in record 1986-16215-001.) Describes a means for determining circumstances when ordinary least squares/moderated multiple regression (OLS/MMR) may be at risk in moderator applications and suggests an alternative regression procedure that can be used to overcome the threat of Type II error posed by these circumstances. Using field study data on job satisfaction of employees at state institutions for the developmentally disabled, parallel analyses are presented to show how markedly different results can be obtained between OLS/MMR and the remedial procedure. It is suggested that investigators who have reached conclusions with the traditional OLS/MMR approach reexamine their data. (23 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Failure to detect moderating effects: Is multicollinearity the problem?

We show that Morris, Sherman, and Mansfield's (1986) contention that multicollinearity causes ordinary least squares–moderated multiple regression (OLS–MMR) to underestimate the importance of moderator effects is incorrect for their own data. Multicollinearity was reduced to the point that it was negligible by transforming predictor variables and moderator variables to standard scores prior to computing cross-product terms. We show the resulting cross-product terms both mathematically and empirically to have near-zero correlations with standardized predictors and moderators. Yet, as Arnold and Evans (1979) showed, the results of OLS–MMR are unchanged by this linear transformation of scale. Morris et al's (1986) finding of significant moderator effects when using principal-components regression (PCR) is probably a result of some artifact of PCR. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Some problems with the structural model and a remedy.

Although the structural model continues to be the major paradigm for psychoanalytic theory, it is flawed by major problems. Among them are the uncertain relationship of the dynamic unconscious to the id, ego, and superego; whether these macrostructures are to be conceptualized as classifications of functions or as objectively existing causally effective entities; and difficulties with how the model accounts for the interaction between external reality, unconscious fantasy, and the psychic apparatus. The remedy offered is a model wedded to the dynamic and genetic points of view that resolves these problems, accommodates such concepts as trauma, repetition, and sublimation, and dispenses with highly abstract concepts. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Relative power of moderated multiple regression and the comparison of subgroup correlation coefficients for detecting moderating effects.

A Monte Carlo simulation assessed the relative power of 2 techniques that are commonly used to test for moderating effects. 500 samples were drawn from simulation-based populations for each of 81 conditions in a design that varied sample size, the reliabilities of 2 predictor variables (1 of which was the moderator variable), and the magnitude of the moderating effect. The null hypothesis of no interaction effect was tested by using moderated multiple regression (MMR). Each sample was then successively polychotomized into 2, 3, 4, 6, and 8 subgroups, and the equality of the subgroup-based correlation coefficients (SCC) was tested. Results show MMR to be more powerful than the SCC strategy for virtually all of the 81 conditions. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Distinguishing between moderator and quadratic effects in multiple regression.

Moderated regression analysis is commonly used to test for multiplicative influences of independent variables in regression models. D. Lubinski and L. G. Humphreys (1990) have shown that significant moderator effects can exist even when stronger quadratic effects are present. They recommend comparing effect sizes associated with both effect types and selecting the model that yields the strongest effect. The authors show that this procedure of comparing effect sizes is biased in favor of the moderated model when multicollinearity is high because of the differential reliability of the quadratic and multiplicative terms in the regression models. Fortunately, levels of multicollinearity under which this bias is most problematic may be outside the range encountered in many empirical studies. The authors discuss causes and implications of this phenomenon as well as alternative procedures for evaluating structural relationships among variables. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Coping with Gulf War combat stress: Mediating and moderating effects.

The authors examined relationships between method of coping with combat-related stress and psychological symptoms among Gulf War Army personnel (N?=?1,058). Participants were surveyed on return from the Gulf region (Time 1) with the Coping Responses Inventory (R. Moos, 1990) and a measure of combat exposure. Outcomes were symptom measures of posttraumatic stress disorder (PTSD) and depression. At Time 2 (18–24 months) participants completed the same symptom measures and an index of postwar stress. Higher proportions of approach-based coping in the war zone were related to lower levels of psychological symptoms. Combat exposure moderated the effects of coping on Time 1 PTSD. Coping predicted changes in symptoms of depression but not PTSD. Combat exposure affected changes in depression through postwar stress but had a direct negative effect on PTSD. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

An analysis of situational moderating effects on the relationships between least preferred co-worker and other psychological measures.

Correlations were computed between a manager's least-preferred-coworker (LPC) score and 17 psychological measures and moderated by subordinates' satisfaction with the manager in 34 offices of a nationwide brokerage firm. Results tend to support the hypothesis that correlations between LPC and the psychological measures should be opposite in direction for high- vs low-satisfaction offices. The hypothesis was directionally confirmed for 11 of 17 pairs of correlations, and the differences were statistically significant for 5 of the 17 pairs. Results also support F. E. Fiedler's interpretations of effective leader behavior and the relation between LPC and behavior. (37 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Hedonic reactions to the disconfirmation of product performance expectations: Some moderating conditions.

It is argued that affective judgments of a hedonic nature whereby individuals react adversely to an unexpected positive as well as a negative event have been observed on an infrequent basis, primarily because the facilitating conditions have not been specified. Hedonic affective reactions to the disconfirmation of expectations of a recently introduced automobile were examined under 3 conditions representing ego involvement, attitudinal commitment, and task interest. Correlations between degree of absolute disconfirmation and affective judgments made by 488 undergraduates were calculated for 8 treatment combinations. Hedonic reactions were observed only when involvement, commitment, and interest were high. A separate chi-square analysis of the correlation z transformations showed that two-thirds of the cell variance was due to treatment main effects. (16 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Work stress and self-reported alcohol use: The moderating role of escapist reasons for drinking.

This study examines the moderating role of escapist reasons for drinking alcohol in the job stress/self-reported alcohol use and problems relationship. It was hypothesized that higher levels of job stress would be associated with higher levels of self-reported drinking (H1) and drinking problems (H2) only for those who endorsed escapist reasons for drinking. For those who did not hold such beliefs, higher levels of job stress were predicted to be associated with lower self-reported alcohol intake (H3) and problems (H4). Survey data from white- and blue-collar workers employed across all paycodes and positions were collected randomly at a large manufacturing organization (62% response rate). Participants responded to questions concerning work stress, reasons for drinking, alcohol intake, and alcohol problems. Using only nonabstainers with complete data (N?=?1,645), results from regression analyses generally supported all hypotheses. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Inference in regression models of heavily skewed alcohol use data: A comparison of ordinary least squares, generalized linear models, and bootstrap resampling.

Analysis of alcohol use data and other low base rate risk behaviors using ordinary least squares regression models can be problematic. This article presents 2 alternative statistical approaches, generalized linear models and bootstrapping, that may be more appropriate for such data. First, the basic theory behind the approaches is presented. Then, using a data set of alcohol use behaviors and consequences, results based on these approaches are contrasted with the results from ordinary least squares regression. The less traditional approaches consistently demonstrated better fit with model assumptions, as demonstrated by graphical analysis of residuals, and identified more significant variables potentially resulting in theoretically different interpretations of the models of alcohol use. In conclusion, these models show significant promise for furthering the understanding of alcohol-related behaviors. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Methodological artifacts in moderated multiple regression and their effects on statistical power.

Monte Carlo simulations were conducted to examine the degree to which the statistical power of moderated multiple regression (MMR) to detect the effects of a dichotomous moderator variable was affected by the main and interactive effects of (a) predictor variable range restriction, (b) total sample size, (c) sample sizes for 2 moderator variable-based subgroups, (d) predictor variable intercorrelation, and (e) magnitude of the moderating effect. Results showed that the main and interactive influences of these variables may have profound effects on power. Thus, future attempts to detect moderating effects with MMR should consider the power implications of both the main and interactive effects of the variables assessed in the present study. Otherwise, even moderating effects of substantial magnitude may go undetected. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Failure to complete treatment for headache: A multiple regression analysis.

Used a hierarchical multiple regression analysis to predict the number of sessions completed by 90 patients in a headache-treatment program. Demographic variables accounted for 18% of the variance (p?p?     

Norming clinical questionnaires with multiple regression: The Pain Cognition List.

Questionnaires for measuring patients' feelings or beliefs are commonly used in clinical settings for diagnostic purposes, clinical decision making, or treatment evaluation. Raw scores of a patient can be evaluated by comparing them with norms based on a reference population. Using the Pain Cognition List (PCL-2003) as an example, this article shows how clinical questionnaires can be normed with multiple regression of raw scores on demographic and other patient variables. Compared with traditional norm tables for subgroups based on age or gender, this approach offers 2 advantages. First, multiple regression allows determination of which patient variables are relevant to the norming and which are not (validity). Second, by using information from the entire sample, multiple regression leads to continuous and more stable norms for any subgroup defined in terms of prognostic variables (reliability). (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Support behavior and relationship satisfaction in couples dealing with diabetes: Main and moderating effects.

This study examined associations between support behavior, i.e. active engagement and protective buffering, and relationship satisfaction in both patients with diabetes and their partners. Active engagement refers to supportive behavior characterized by involving one's partner in discussions, asking how the other feels, and problem solving strategies. Protective buffering refers to less supportive behavior characterized by denying fears and worries, and by pretending everything is fine. Furthermore, we examined whether there were interactive effects of these two support behaviors on patients' and partners' relationship satisfaction. At baseline (T1), 205 couples rated to which degree they received active engagement and protective buffering from their partners, and completed a measure of relationship satisfaction. At three follow-up assessments, couples were asked to fill out the same measures. Using dyadic data analytic approaches, we found relationship satisfaction to be positively associated with active engagement, and negatively with protective buffering, in both patients and partners. Moreover, we found a moderating effect, in that the negative association between protective buffering and relationship satisfaction was only present when levels of active engagement were relatively low. Again, these results were found for patients as well as their partners. We were able to replicate the T1 results at the other three assessment points. Our findings illustrate the need to consider adequate and less adequate support behaviors simultaneously, and to study the effects on both patients and partners. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

The predictive power of ridge regression: Some quasi-simulation results: Correction to Cattin.

Reports an error in the original article by Philippe Cattin (Journal of Applied Psychology, 1981, Vol. 66, No. 3, pp. 282-290). The fifth sentence in the first paragraph on page 284 contains an error. The sentence should read: "For intermediate values of k?, the ridge regression weights are 'weighted sums' of the OLS regression weights [not models] and of the zero-order sample correlations and tend to decrease in absolute value as k? increases." (The following abstract of this article originally appeared in record 1981-27117-001.) Reviews research indicating that ridge regression tends to improve the mean square error of prediction obtained with ordinary least squares (OLS) regression in a wide range of conditions. MMPI profiles of 861 psychiatric patients and diagnoses from 29 psychologists, as collected by P. E. Meehl (see record 1960-04396-001), were used to examine the gains in cross-validated multiple correlation obtained with ridge regression compared with OLS and equal weights as a function of sample size and ridge constant. A simple formula for estimating the ridge constant was also evaluated. Results that are related to recent developments concerning the use of Bayesian regression procedures show that ridge regression improves both the mean square error of prediction and the cross-validated multiple correlation obtained with OLS when the ratio sample size to number of predictor variables is relatively small. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Some effects of rejection upon attraction to a group.

Adistinction was made between invidious rejection, defined as rejection for reasons that are likely to reflect upon the Ss' self-esteem, and noninvidious rejection. On the basis of the hypothesis that invidious rejection would arouse a need for social reassurance, it was predicted that: (a) for any given degree of rejection, loss of attraction to the group will be smaller when the reason for rejection is invidious than when it is noninvidious, and (b) that the differences in loss of attraction between invidious and noninvidious conditions will be greater after strong than after mild rejection. The results with respect to attraction to the group generally supported Hypothesis 1, but not Hypothesis 2. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Faculty ratings and student grades: A university-wide multiple regression analysis.

Studied the relationship between grades professors give their students and ratings those students give their professors, using multivariate techniques and a large sample size (2,360 of 2,449 course sections taught in the spring semester of 1973) to avoid weaknesses of previous studies. Results show the following: (a) Only one factor was present among the 8 rating items; (b) the correlation between average student grade in each course section and average student rating of the teacher of that section was .35; (c) average grade was the best predictor of average rating; and (d) when average grade was added to several other available predictors, it significantly improved the multiple correlation from .25 to .39. Findings suggest that students' grades probably do influence their ratings of faculty, accounting for about 9% of the total variance. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Rhymes and reasons: Some aspects of the child's appreciation of poetic form.

Investigated 18 1st and 3rd graders' appreciation of rhyme, rhythm, and alliteration in poetry. Several measures were used to assess each child's appreciation of the poetic devices. Even 1st graders exhibited some appreciation of rhyme and rhythm in poetry. These Ss displayed an ability to attend to these poetic devices as measured by a detection task, 2 concept learning tasks, and a production task. Moreover, they also expressed preferences for poetry samples containing these devices. However, such was not the case for alliteration. First graders experienced difficulty in attending to alliteration, and neither 1st nor 3rd graders displayed any tendency to prefer poems with alliteration. The results of a posttest questionnaire indicated that only a few children showed any understanding of how the devices function in poetry. It is concluded that although 1st graders have some rudimentary appreciation of rhyme and rhythm in poetry, the ability to understand how such devices function in poetry is a much later development. (18 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Employee reactions to ergonomic job design: The moderating effects of health locus of control and self-efficacy.

A field survey of 180 municipal government office employees (82% women, 21–75 years old) investigated the potential moderating effects of internal health locus of control (HLOC) and self-efficacy on employees' reactions to ergonomic job design. Internal HLOC moderated the associations between ergonomic job design and somatic complaints and turnover intentions, and, to a lesser extent, job satisfaction. Self-efficacy moderated the associations between job design and job satisfaction, somatic complaints, and, to a lesser extent, persistent pain. Employees with low self-efficacy or low internal HLOC were influenced more by their physical job conditions than those with high self-efficacy or high internal HLOC. Implications for the ergonomic design of offices are discussed. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Parametric statistics and levels of measurement: Factorial designs and multiple regression.

Let Y be a continuous, ordinal measure of a latent variable Θ. In general, for factorial designs, an analysis of variance of the observed variable Y cannot be used to draw inferences about main effects and interactions on the latent variable Θ even when the standard normality and equality of variance assumptions hold. If Y is a continuous, ordinal measure of a latent variable Θ; X?,…, Xn are continuous, ordinal measures of latent variables Φ?,…, Φn; and the observed measures have a multivariate normal distribution, then a multiple regression analysis of the observed criterion measure Y and predictors X?,…, Xn can be used to test hypotheses about multivariate associations among the latent variables. Furthermore, the predicted values Y′ are unbiased estimates of quantities that are monotonically related to predicted values on the latent criterion variable Θ. (PsycINFO Database Record (c) 2010 APA, all rights reserved)