So, I feel like what I've had in mind of statistics is sort of fragmented, but have trouble putting pieces together in a coherent and systematic manner. If I'm asked about categorical outcome variables, I can talk a bit about it. If I hear some conversation about fixed and random effects, I can chime in a bit as well. The same happens with multi-level models, OLS violations, and all those sorts of things.
I think what I need now is a framework, or a book, that helps to think about the strategy of data analysis. It does not necessarily be some thing demonstrating analysis technique or error structure formula. Rather, it should be an instruction of what to think when exploring data, what to consider when constructing models, decision trees of models, etc.
One (just one) piece of insights I'm looking for looks like these comments/critique of repeated (M)ANOVA:
- They assume categorical predictors.
- They do not handle time-dependent covariates (predictors measured over time).
- They assume everyone is measured at the same time (time is categorical) and at equally spaced time intervals.
- You don’t get parameter estimates (just p-values)
- Missing data must be imputed.
- They require restrictive assumptions about the correlation structure.
If you know books or sources of material that can potentially tackle my problem, please suggest. I don't know if I say it clear enough ('cause I'm myself confused now ) But I feel it's like a medical doctor facing a patient. The doctor knows the patient has some problem, and he knows some medications. What he needs to know is how to, step by step, do an appropriate diagnosis. Failing that, the prescription would be dysfunctional at best and fatal at worst.
Last edited by non-sleeper; 11-01-2012 at 05:00 PM.
In God we trust; all others must bring data - Deming
Advertise on Talk Stats