I wonder if anyone can offer me some advice on the most appropriate model to use with my data set, it has several unusual aspects that prevent me from using similar models (MANOVA etc.) as far as I can tell. It is rather difficult to explain the problem, but I will try to outline it as best I can (without coming from a Statistics background).


First I have a group of subjects, that are measured under 3 conditions.


For each subject measurement is taken in each condition. The measurement consists of several results that will be detected gradually over different times (however it cannot be predicted when these times will occur, therefore there will be different time points across subjects/conditions) and each result consists of two weakly correlated dependent variables.




There are a few questions I would like to answer with statistical analysis:


1) The between subject variance is significantly greater than the within subject effects of condition, is there a way to control for this? Would a MANOVA with some covariant be sufficient?


2) Can the changes in the slope of the line be predicted/explained by other factors (covariants)?


3) Are the changes in one of the dependent variables due to the effect of condition or due to a change in the other dependent variable being measured?


4) A more general question I would like to know is how the slopes change with condition. If a MANOVA is suitable to test the effects of condition is there any post hoc tests that I could do to show this?




I think regression analysis could be used to look at the data within one condition, but I'm having trouble knowing how to incorporate the repeated measures/different conditions of the data. I have plotted the two dependent variables as a function of the time of detection for each subject. (To further complicate things one of the dependent variables has a skewed-right distribution, is it better to log-transform this data before using regression analysis?) So each subject will have two regression lines associated with them per condition.


Any advice would be much appreciated, I'm very new to statistics and don't really know where to take my analysis from here!