InDiMa SIP Group

3. STAGE 3: MULTIPLE REGRESSION MODELING

3.1       ‘I-D-E-A’ INTERACTION & COMMUNICATION STYLES (PART 3 OF QUESTIONNAIRE)

The four ’I-D-E-A’ Interaction & Communication (factor) scale scores for the sample are presented in Table 1 (see page 4): Psychometric Characteristics of the Scales (n = 1000).

3.2       FOUR FACTOR ICS AND I-D-E-A MULTIPLE REGRESSION (PART 1 ‘PERSONAL PSYCHOLOGICAL PROFILE’ AND PART 2 ‘SOCIAL ENVIRONMENT PROFILE’)

Multiple regression models were used to simultaneously examine the relationship of demographics, clinical characteristics and communication styles with the four factor scores of the ‘Personal and Social Profile’ (23 scales) and the I-D-E-A (4 scales) created from the – in total – 27 scales.

 

3.2.0   Statistic Significance (Table 7) and Relevance

Indicators that were statistically significant in the presence of the other variables are noted in the Table 7.

The standardized betas are presented. They are standardized by type of measurement so that the relative importance of the relationships can be assessed. Larger coefficients, either negative or positive, imply a greater relationship with the factor.

Due to the large number of tests and large samples size, a correction to the error probability should be made. We interpret as statistically significant, those p values of .001 (***) or less.

However, for illustrative purposes all results with p < .05 will be discussed when describing the regression models. The statistical relevance is determined by the ‘variance explained’, i.e., correlations above r = .33 (explaining more than 10% of the variance) might be considered as ‘relevant’.

TABLE 7: STANDARDIZED REGRESSION COEFFICIENTS FOR THE ASSOCIATION WITH THE FOUR FACTORS

Indicators Factor I Factor II Factor III Factor IV
Demographics
Age .12*** -.22*** .02 .14***
Gender .05 -.04 .08** -.13***
White / Caucasian .04 -.02 .13*** .01
At least some college -.05 -.02 .04 -.08**
Married .05 -.01 .05 .04
Clinical Diabetes Characteristics
Duration of Diabetes .03 -.04 -.04 .09**
Type I Diabetes .12*** .07 -.03 -.11**
Treatment Intensity -.10** .02 .05 -.03
Insulin prescribed at first diagnosis .07* .01 .05 .05
BMI1) -.31*** .02 .05 .03
I-D-E-A Interaction & Communication Styles
Social Initiative (Dimension I) .21*** .04 -.03 .08*
Social Style (Dimension II) .02 -.14*** .20*** .14***
Social Flexibility (Dimension III) .09* .06 .07 .05
Self-Steering (Dimension IV) .05 -.05 .01 .04
Multiple Correlations (R2) for the Sections and Statistical Significance
Demographics .02** .08*** .03*** .06***
Clinical diabetes characteristics .15*** .01 .01 .01*
I-D-E-A .06*** .02** .05*** .06***
Total Variance Accounted for 23% 11% 9% 13%

1)  BMI = Body Mass Index

1 Mean used for missing; * p < .05; ** p < .01; *** p < .001 General Findings

The results of the regression modeling show that in total, the demographic, and clinical and interaction and communication style indicators explained significant variance in all four factor scores.

The influence, however, is only ‘relevant’ (explaining more than 10% of the variance) for the interrelation of (R2 = .15***) ‘(Realistic) Self Evaluation’ and ‘Coping with Diabetes’ with being type I diabetic, with ‘first described insulin’, with ‘intensity of treatment’, and with BMI.

3.2.1      Factor I: ‘(Realistic) Self Evaluation and Coping with Diabetes

About a quarter of the Factor I variance (23%) is explained by age, having type I diabetes, the intensity of treatment, if insulin was first prescribed and BMI.

Individuals who were older, had type 1 diabetes, have medical treatment for their diabetes, and were prescribed insulin at their first diagnosis, as well as individuals with lower BMI had higher Factor I ‘(Realistic) Self Evaluation and Coping with Diabetes’ scores.

In addition individuals with higher social initiative and social flexibility had higher Factor I scores.

 

3.2.2      Factor II: ‘Negative External Factors and Tipping Points’

Demographic and I-D-E-A scores, but no clinical diabetes factors explained significant variance in Factor II.

The model explained 11% of the variance and had only two significant indicators:  age and ‘Social Style’ (I-D-E-A) in communication.

Younger individuals and those with a reserved and introspective style of communication had higher Factor II scores, i.e., were subject of a negative influence of the ‘social environment’, remaining in a state of ‘denial’, not coping with the ‘reality’ of diabetes type II.

 

3.2.3   ‘Accept Clarity and Guidance: Willingness to Learn’

90% of the variance in Factor III was explained by demographics and IDEA style indicators.

Gender and race, along with a social style of communication were the significant indicators in the model:

White / Caucasian females were more likely to have higher factor III scores, i.e., willingness to learn, as were those with a ‘Social Style’ in communication which are ‘open, emotional, and intuitive’ rather than ‘controlled and rational’.

 

3.2.4   ‘Adaptive Living with Type II Diabetes’

Demographic, clinical, and ‘I-D-E-A Interaction & Communication Styles’ explained 13% of the variance in Factor IV.

Seven variables were statistically significant:

Individuals with higher scores on Factor IV were more likely to be older, male, less educated, have type II diabetes, have had it for a long duration, and have higher social initiative (pro-active, dominating) and an open, intuitive social style in communication.

3.3          EXTERNAL VALIDITY OF THE FOUR ICS FACTORS

The external validity of the factors is good:

  • Demographic characteristics explain between 2% and 8% of the variance in the factors.
  • Evaluation: This is very good, as it shows that the factor scores and not biased extensively by demographic characteristics. Because there are associations, most particularly with age and gender, Prof. Dr. Joan Russo advises for age, gender, ethnicity and education to be used as ‘covariates’ or ‘factors for adjustment’ in subsequent statistical analyses using these factors.
  • The clinical diabetes characteristics were strongly related to Factor I ‘(Realistic) Self Evaluation and Coping with Diabetes’ (15%), weakly related to Factor IV (1%) and not significantly related to factors II and III. We consider the 1% explained variance with Factor IV is irrelevant.
  • Evaluation: This shows that the Factor I scores do reflect characteristics of the disease. These finding will be used as covariates in any future modeling.
  • The ‘I-D-E-A Interaction & Communication Styles’ were significantly related to all four Factors. In general, the ‘Social Style’ (open and extrovert) was related to the four factors explaining between 2% and 6% of the variance in the factors.
  • Evaluation: This shows that although the concepts of communication and the factors are somewhat similar, they are relatively independent, sharing less than 10% common variance.
  • This is very good, and implies that that I-D-E-A Interaction & Communication Style scales and the Factor Score scales can be used independently (‘side by side’) in models offering ‘options’ for action or predicting outcomes.

Of course, demographic and clinical characteristics would have to be used as covariates.