Statistics for Research - L29 - Bootstrap Independent Samples T Test using SPSS

Bootstrapped independent-samples t testing in SPSS provides a way to compare two groups on a mean-based outcome even when the data are not normally distributed. In this case, the outcome is “Vision” (communication and perception of vision), and the groups are male versus female respondents. Because normality is not assumed, the analysis uses bootstrapping rather than relying on the standard parametric t test assumptions.

The workflow starts with Analyze → Compare Means → Independent Samples T Test, then assigns “Vision” as the test variable and “Gender” as the grouping variable (coded as GR1 and GR2). To implement bootstrapping, the options are set to draw 5,000 bootstrap samples and to use a bias-corrected and accelerated (BCa) confidence interval type, which is designed to stabilize inference through resampling. Effect size estimation is also requested to obtain Cohen’s d values (labeled as “Cohen’s d” in the output).

The results begin with group statistics. Male respondents (n = 288) have a mean of 5.08, while female respondents have a mean of 4.73 on a 7-point scale. Both groups sit in a similar range, suggesting some agreement that vision is communicated effectively within the organization, but the key question is whether the difference between means is statistically meaningful.

To decide how to interpret the t-test output, the analysis checks Levene’s test for equality of variances. Levene’s test is not significant here, meaning equal variances can be assumed. That matters because SPSS reports separate rows for “equal variances assumed” and “equal variances not assumed,” and the correct row should align with the variance assumption.

Using the bootstrapped inference aligned with “equal variances assumed,” the bootstrap significance value is 0.011, which is treated as not significant in the session’s interpretation (compared against a 0.05 threshold). The confidence interval spans zero (with values shown as negative to positive around 0), reinforcing that the observed mean difference is not statistically distinguishable from sampling noise. The point estimate for Cohen’s d is also small, indicating only a minimal practical difference between male and female perceptions of vision communication.

Overall, the analysis concludes that male and female respondents do not differ significantly in their perceptions of how well organizational vision is communicated, with both the bootstrapped significance test and the confidence interval pattern pointing to an absence of a reliable mean difference. The session’s practical takeaway is the SPSS procedure: set up the independent-samples t test, enable bootstrapping with 5,000 resamples and BCa intervals, check Levene’s test to choose the appropriate variance row, and interpret significance and confidence intervals around zero to judge group differences.