Dear all,
I need to run an analysis and have tried some (non-satisfactory) approaches, so I would be most grateful if someone could help me with the following problem: I have measured a given response (Z) along many years (Years = 6) in various places (P=35). That is, I have a total of 210 observations [6 years * 35 places]. I also have data on a number of possible factors (some 35 in total, measured for each year in each place) that could be related to my response. So I would like to run a 'stepwise regression' sort of analysis so that I could explore which factors would better explain the variance in my response (Z). However, I would not like to loose the information I have for every year, e.g. by averaging all years for one place (as the average here would not make sense). I wonder if there would be any sort of analysis of this type (response and predictors are continuous & predictors are selected by the model) where the analysis is nested by subject (in my case 'place' would be the equivalent of 'subject'). I have heard of possibilities such as 'repeated-measures multiple regression' and 'repeated-measures ancova', but it seems that in these analyses all variables will be part of the final model (and I have far too many, and do not know which ones are important - this is actually what I want to know with the model). Any clues here would be really of great help (to run the statistical analyses I have spss and matlab).
Thanks a lot and best wishes
Cynthia
email: [log in to unmask]
|