If things are significant at .001, then there is probably an interaction, but you should also do a correction for multiple comparisons.

doug

On 05/19/2012 09:55 PM, ZhiLiangLong wrote:

Hi Doug: Thank so much for your suggestions. Please find Freesufer Digest, Vol 99, Issue 23, Topic 11: Question about DODS model and DOSS model. As you suggested, I have four regressors: Group1 Intercept, G2 Intercept, G1 slope, G2 slope, and the contrast [0 0 1 -1].I will test the interactions between group and age. Is that to say to test the null hypothesis that G1 slope is equal with G2 slope? In qdec, while the analysis was complete, I cliked the Display tab and saw a list of questions summarizing the various analyses that were completed. And i cliked on 'Does the thickness--age correlation differ between Male and Female?' to test whether there is no interactions, and the qdec displayed the statistically significant regions. I set threshold to 0.001 and found all regions showing decreased thickness-age correlation in Male relative to Female(no increased thickness-age correlation was found). So, may i draw the conclusion that there is no interactions between group and age? However, if i set the threshold lower (e.g. 0.01), both increased and decreased thickness-age correlation was found, does that mean the interactions exist? if so, how do i set the threshold? Hope you can give me some suggestions.