Hello everyone,
I am quite confused on whether this is the correct step to take towards analysing my data. I will try to briefly explain what I mean by the question. I am using a questionnaire that tests for levels of religiosity using five factors. This was tested on a Christian and a Muslim group. I want to know which of these five factors best predict attitudes toward counselling for Christians alone and Muslims alone. So my question is would it be appropriate to run a regression analysis using the Muslim group and then run it again using the Christian group only?
yes just exactly you say. Firstly, you have to split your file with respect to religiosity variable. It is easliiy accomplished by SPSS-data-Split File-click organize output by groups then send your religiosity variable that side then click okey.
In the second stage ıf you want to compare your models predictive validity (e.g Is my model better fit for christians than muslims?) you have to conduct a R square difference test which easliy conduct in FZT program which is annex. use your R values in your regression model summaries and use Fisher Z in FZT program.
If you want to compare your individual beta coeeficients in multiple regression (e.g ıs my self esteem variable more strongly predict religious involvement in muslims than christians?). It is a bit complex but you can find more information on appendix 2 (rhtest2a), and https://stats.idre.ucla.edu/stata/faq/how-can-i-compare-regression-coefficients-across-3-or-more-groups/ and http://psych.unl.edu/psycrs/statpage/regression.html
Yes, as Professor Sahin wrote, you would run the regression analysis for each of the groups separately. One consideration would be the resulting sample size in each of those two groups. If the resulting group sizes are smaller than 30-40, or simply very different from each other or if the variances in the two groups were very different, you may need to be cautious in comparing the two results. On the other hand, if your total number of respondents is much higher (more than 250-300), you could run a structural equation model that would compare the fit of your model for the two groups in one analysis. This would be a complex process and I would not necessarily recommend it, but it might be worth commenting on this option when providing the rationale for the decisions you made in choosing your analyses.
As a religionist who knows something about statistics, it would be very valuable to my research if the regression analysis were run for only the Muslim group and for only the Christian group.
- Badges
- Science topic