Results
The following analyses were conducted for the purposes of examining the possibility of gender
related construct bias in the Behavioural Inhibit Scale (BIS) and in the Humour Styles
Questionnaire (HSQ). Additionally, a gender related predictive bias analysis was also conducted.
Specifically, the HSQ composite scores were regressed onto the BIS composite scores for both
the males and females.
Construct Bias: Behavioural Inhibition Scale (BIS)
In order to test for the possibility of construct bias in the BIS, a scree plot analysis was
conducted on the female and male portions of the sample. As can be seen in Figure 1, the scree
plots for both the males and females were very similar in that they both indicated the extraction
of a single component. Thus, based on this analysis, there was no suggestion of construct bias.
Next, a principal components analysis (PCA) was performed on the female portion of the
sample. Prior to conducting the PCA, the correlations between the BIS items were examined
under the expectation that they would evidence a pattern of positive associations. As can be seen
in Table 1, the correlations were mostly positive and were associated with a M = .17 and SD =
.18. In accordance with the scree plot results, a single component was extracted from the female
correlation matrix. As can be seen in Table 1, item 2, item 3, and item 9 were associated with
communalities less than the minimum expectation of .09. As the BIS is an established measure,
these items were not removed from the analysis. The component was associated with an
eigenvalue of 3.23 and accounted for 29.3% of the total variance. As can be seen in Table 1, item
3 and item 9 were associated with component loadings equal to .02, and .19, respectively, which
is lower than the minimum expectation of .20. Finally, the female BIS composite scores were
associated with a Cronbach’s α = .69, which is very near the minimum level required for
exploratory research (i.e., .70; Nunnally & Bernstein, 1994).
The correlations between the items in the male portion of the sample are reported in Table 2. As
per the female portion of the sample, the correlations were mostly positive and were associated
with a M = .18 and a SD = .19. In accordance with the scree plot results, a single component was
extracted from the male correlation matrix. As can be seen in Table 2, item 9 and item 11 were
associated with communalities less than the minimum expectation of .09. As the BIS is an
established measure, these items were not removed from the analysis. The component was
associated with an eigenvalue of 3.34 and accounted for 30.3% of the total variance. As can be
seen in Table 2, item 9 was associated with a component loading equal to -.11, which is lower
than the minimum expectation of .20. Finally, the male BIS composite scores were associated
with a Cronbach’s α = .70, which is at the minimum level required for exploratory research (i.e.,
.70; Nunnally & Bernstein, 1994).
Based on a Pearson correlation between the female and male component solutions, a factor
congruence coefficient was estimated at r = .78. Thus, based on the guidelines provided in
PSYC3302, one would conclude that there is at best weak evidence to suggest the presence of
construct bias associated with the BIS.
Construct Bias: Humour Styles Questionnaire (HSQ)
In order to test for the possibility of construct bias in the HSQ, a scree plot analysis was
conducted on the female and male portions of the sample. As can be seen in Figure 2, the scree
plots for both the males and females were very similar in that they both indicated the extraction
of a single component. Thus, based on this analysis, there was no suggestion of construct bias.
Next, a principal components analysis (PCA) was performed on the female portion of the
sample. Prior to conducting the PCA, the correlations between the HSQ items were examined
under the expectation that they would evidence a pattern of positive associations. As can be seen
in Table 3, the correlations were all positive and were associated with a M = .36 and SD = .13. In
accordance with the scree plot results, a single component was extracted from the female
correlation matrix. As can be seen in Table 3, all of the items were associated with
communalities equal to or greater than the minimum expectation of .09. The component was
associated with an eigenvalue of 3.57 and accounted for 44.6% of the total variance. As can be
seen in Table 3, all of the items were associated with component loadings equal to or greater than
the minimum expectation of .20. Finally, the female HSQ composite scores were associated with
a Cronbach’s α = .82, which exceeds the minimum level required for exploratory research (i.e.,
.70; Nunnally & Bernstein, 1994). The correlations between the items in the male portion of the
sample are reported in Table 4. As per the female portion of the sample, the correlations were all
positive and were associated with a M = .32 and a SD = .12. In accordance with the scree plot
results, a single component was extracted from the female correlation matrix. As can be seen in
Table 4, all of the items were associated with communalities equal to or greater than the
minimum expectation of .09. The component was associated with an eigenvalue of 3.34 and
accounted for 41.7% of the total variance. As can be seen in Table 4, all of the items were
associated with component loadings equal to or greater than the minimum expectation of .20.
Finally, the male HSQ composite scores were associated with a Cronbach’s α = .79, which
exceeds the minimum level required for exploratory research (i.e., .70; Nunnally & Bernstein,
1994).
Based on a Pearson correlation between the female and male component solutions, a factor
congruence coefficient was estimated at r = .25. Thus, based on the guidelines provided in
PSYC3302, one would conclude that there is strong evidence to suggest the presence of
construct bias associated with the HSQ.
Predictive Bias: BIS as a Predictor of HSQ
In order to test the hypothesis that the BIS composite scores may be associated with predictive
bias, the HSQ composite scores were regressed onto the BIS composite scores separately for
both females and males. Both intercept bias and slope bias were examined using 84% confidence
intervals. That is, bias would be indicated in the instance where the 84% confidence intervals do
not intersect.
As can be seen in Table 5, the female and male intercepts were estimated at 31.77 and 18.80,
respectively. As the upper bound (24.21) of the smaller estimate (i.e., males) did not intersect
with the lower bound (24.88) of the larger estimate (i.e., females), there was evidence for
predictive bias with respect to the intercept.
As can be seen in Table 5, the female and male slopes were estimated at .02 and .52,
respectively. As the upper bound (.32) of the smaller estimate (i.e., females) did intersect with
the lower bound (.28) of the larger estimate (i.e., males), there was no evidence for predictive
bias with respect to the slopes.