Complicating Things
I just finished an initial draft proposal for two studies, each with two hypotheses. My theory is pretty strong but I’m very much still working on my methods; I may change my predictor variables and I need to elaborate on my analysis methods A LOT. The thing that kills me about this proposal is that I had to choose hypotheses that require relatively obscure methods that I’ve never used before. Two of the hypotheses use multilevel structural equation modeling, which isn’t that far outside my knowledge zone since I know multilevel modeling and structural equation modeling, but my sample size will be sup par and the coding for these analyses have the potential to be complicated. For the other two hypotheses I’m currently proposing to use univariate and multivariate tobit regression since my outcome variables are ratios and there is (theoretical) bottom censoring. Tobit regression is used in criminology and econometrics but it’s rarely used in public health. Likewise public health tends to neglect multivariate regression in favor of just running separate models. I could get at similar concepts to what I’m interested in using different analysis techniques for slightly different hypotheses, but of course I’m stuck on the current hypotheses I have that require ratio outcome variables. I am not satisfied with multivariate tobit regression for my study one hypotheses. I want to more examine the relationship between two dependent variables, both of which are ratios, more directly than multivariate analysis may permit. Why do I make things complicated for myself? Also, does anyone search or follow analysis tags on tumblr besides me? I think I know the answer to that.... does this mean that I need to find myself a life?
