You could first check if the manipulation had a significant effect. Therefore, run a factorial 2 x 6 repeated measures anova (Between x Within; control vs. intervetion Group and 6 Weeks of repeated measures of step counts) on the step counts. There shoud be a main effect of Group and an interaction with Weeks. In the intervention Group, there should/could be a 1rst order polynomial trend, i.e. a linear increase in number of step counts.

Secondly, you could calculate the correlations between step counts and questionnaire responses. You say they are nominal. I suppose they are Likert type responses. The correlations should be about the same under all conditions. One can test for differences in correlations online:

https://www.psychometrica.de/correlation.html

Finally, you could run the same 2 x 6 repeated measures anova on the questionnaire responses. If these are not normally distributed (Shapiro test & inspection of histograms and QQ-plots) you can relay on non-parametric tests such as the ARTool:

http://depts.washington.edu/madlab/proj/art/

Here too, there should be a main effect of Group, an interaction with Weeks, and maybe a linear trend within the intervention Group.

In this case, the mental or the behavioral changes you are observing are really dependent on the physical activity i.e. increasing the steps in walking and so on. If you would like to compare two different groups (independent) you can use the unpaired student t-test and the paired student test for the "within- group comparisons". But, however, always first check the pattern of distribution of your data whether normally distributed or not using tests like Kolmogorov-Smirnov test. Normally distributed continuous variables can be compared by t-tests as mentioned above. Otherwise; ManWhitny test has to replace t-test for continuous variables if not normally distributed even after their Log transformation. Chi-Square test is used to compare categorical variables.

Best wishes