Regression Discussion > Which Test to Use?
Dear Sleep Deprived,
To make sure I understand, do you have repeated measures of any of your variables? If so, which variables and how many repeated measures? I ask this because of how you are describing modeling time, which usually indicates repeated measures. If not, are all variables measured at the same time point (cross-sectional)?
Jeremy Taylor
Hi Jeremy,
Thank you for your reply, you are now officially my savior of statistics! There are no repeated measures. 370 participants were given both Bond and Grief surveys to complete and part of the questionnaire asked 'how long ago did your pet die', so people were then categorised into one of the 4 groups, <1yr, 1-2yrs, 3-4yrs or 4+yrs.
I could run a simple regression on each of the 4 time categories and assess the relationship of bond and grief in each distinct group (I have done this and it does answer my hypothesis, i.e. that as bond increases in the first 3 time groups grief decreases, and as bond increases in the last group (4+yr) grief actually increases) but I'm worried this is not the best model to fit my data. What do you think?
Cheers
Still Sleep Deprived
Sleep Deprived
Dear Sleep Deprived,
I think a regression makes sense, though I'd also recommending examining interaction effects along with main effects:
Main effects:
Time
Bond
Interaction:
Bond*Time
If the interaction term is significant, that means the effect of bond on grief varies by Time. I hope that helps!
Jeremy Taylor
I have 2 continuous variables (continuing bonds and grief) and one categorical (time since death - has 4 categories). Grief is my DV and the other 2 are IV's. I'm hypothesizing that time will effect the relationship of bond on grief, in that during the first 3 years since the death of a loved on, as bond increases grief will decrease. And the opposite will occur in the 4th year, i.e. as bond increases grief will also increase.
I'm a bit sleep deprived and am struggling to find the correct method of analysis. I have tried hierarchical multiple regression (and dummy coded the time since death categories) but the outcome doesn't make sense as I know my hypothesis to be true (from exploring the data that I have to work with). I'm considering ANCOVA but this won't tell me anything about the direction or strength of the relationship between bond and grief (dependent on time).
Sorry, this is probably not making much sense! Hopefully you can shed some much needed light on the matter.