October 2008 Poll - Discussion

Strategy for Analyzing Responses

Whole Types

The notion of using a whole-type perspective when analyzing psychological type data has become increasingly emphasized in the type community.1 This whole-type perspective encourages researchers to focus on the four-letter types (e.g., ENFP, ISTJ) as the  primary focus of research in contrast to research that focuses on the four separate preferences (e.g., E-I, S-N).  A key impetus for the whole-type perspective is the notion of synergy--that each of the 16 types is more (or less) than the sum of the individual preferences. Further, various groupings of type preferences often have been used as a way to better understand "subtypes". For example, the combination of S-N and T-F mental functions preferences produces four mental functions groupings--ST, SF, NT, and NF--that have important consequences for life activities such as career choice.

Given this emphasis on type groupings, we decided to analyze the results from the October 2008 poll by focusing on groupings that are of theoretical interest generally (e.g., the 8 Jungian Types) as well as type groupings that have been found in prior research to have some relationship to liberal-conservative political orientations (e.g.. the S-N/J-P groupings). The rationale for selecting each grouping is presented on the following respective pages prior to the discussion of results from the analysis for that grouping.

 

Multivariate Approach

We first analyzed the data using the "Multivariate" option in the General Linear Models procedures provided by SPSS 15.0.  We used the "Multivariate" option given that we have five dependent variables that correlate in varying degrees. (These five variables are discussed on the preceding page.)  This approach is used to control  for experiment-wise error rate (which simply means that the more tests that a researcher conducts, the more likely it is that the researcher will obtain statistically significant results just due to chance, especially if the dependent variables being studied are correlated). Unfortunately, the results from our "Multivariate" analysis suggested the data did not meet certain statistical assumptions. In particular, the assumption of equal variances was not met for most of our political orientations measures. This pattern was stronger for the measures dealing with social perceptions and attitudes; in some analyses for the economic measures, the assumptions were met.

 

Univariate Approach

Given these results, we adopted a univariate approach where we analyzed each of the five dependent variables separately using both oneway ANOVA and the Kruskal-Wallis Test (K-W Test).  As most researchers know, the analysis of variance (ANOVA) approach is a "parametric" approach and thus more powerful statistically than the "non-parametric" K-W Test. To use ANOVA, however, requires that the data meet more restrictive assumptions than is required for the K-W Test. And, as noted above, our data appear to violate some of the assumptions for using ANOVA.

Our strategy thus was to apply both tests. And, in those instances where both tests were significant, we considered the relationship between the type groups and the politics variable to be significant.   To determine which type or type groupings were significantly different, we used the post-hoc test results from the ANOVA as well as the mean ranks from the K-W Test. For the ANOVA post-hoc analyses we selected two tests. The first is the Sidak test which is fairly conservative in that it controls for experiment-wise error rate in the univariate post-hoc test.  This test, however, does require the assumption of equal variances.  In cases where the equal variances assumption were not met, we used Tamhane's T2 test (which also allows for unequal sample sizes in the different type groups).

This univariate approach , of course, interjects concerns about experiment-wise error rate for the whole analysis and thus interpretation of results is based on finding consistent patterns of results.

 

Whole Types or Not?

An article 2 in a recent issue of The Journal of Psychological Type has challenged the assumption of "type dynamics" as used with the MBTI®. The general argument of this article appears to be that the strongest effects of psychological type are likely to result from the individual preference dichotomies individually or in additive combinations--other than those suggested by type dynamics models. This perspective conflicts in many ways with the whole-type perspective.  To address these concerns, we also present an analysis of the data from a more traditional regression approach. 

NOTE:  As we will note at various points in the analyses, the sample size for certain type groupings was insufficient for us to fully address some of the key issues surrounding the whole-type and type dynamics issue.  As might expected, the 4 SP groups were fairly small, thus limiting full consideration of some of the issues raised in the two references below. 


----------

  1Mitchell, W. D. (2006). Validation of the full dynamic model of type.  Journal of Psychological Type, 66(5), 35-48.
  2Reynierse, J. H. (2009, January). The case against type dynamics. Journal of Psychological Type, 69(1), 1-21.

   

 

Print Email

 
 
 
 
 © Politicaltypes.com 2017. All Rights Reserved.

Share This