Dear All I'm having an on-going problem with data analysis for 2 of my studies. I have a number of taste measures (6) and a number of personality measures (12 subscales). Since research in individual differences in terms of personality and taste preference has received limited attention these studies are exploratory in nature. I'm interested in the relationships between taste and personality - why do certain people like certain tastes and not others? Because of the exploratory and correlational design of these studies my supervisor favours multiple regression using the backward method which I have done. I haven't read anything good about this technique but research in the eating behaviour literature does employ this technique (often not very explicit about how they have selected the model of best fit). Unfortunately I do not have enough power to use SEM which would probably be the best option. Other suggestions have included canonical correlations - but I understand that this technique "clumps" DVs together which is not appropriate for my taste measures as these do not intercorrelate indicating hat they should be considered as unique constructs (which they are!). Does anyone have any suggestions or advice? Thanks Catherine Day Sheffield Hallam