Where do I start?

#1
Hi!
I'm feeling overwhelmed here and don't know where to start. It is the second message regarding my project and any additional help would be great.

Basically, my project is on predictors of treatment outcome. I'm looking at pretreatment variables such as symptoms severity, parental control, presence of comorbidity and a bunch of other variables and try to predict treatment response using both presence/absence of diagnosis at posttreatment (in which case I know I need to use logistic regression) but also changes in self-reported symptoms based on questionnaire data (I think MANOVA is appropriate). What I don't know is where to start?

Do I check data for outliers? Normality? Is this done on the independent variables/predictors? Will I also do it for the dependent variables as when the outcome is based on posttreatment seveity scores based of respondents' self-report measured on Likert scale?

I have missing data and my school has no licence for adding missing data in SPSS. With this said, how will I do the above? I understand that doing the analyses in MPlus will take into account missing data without the need to replace them. Is this true? Does MPlus come with syntax that guides one how to conduct the analyses? My advisor said I should do the correlations but did not say what correlations specifically (I need to figure it out on my own - I think it is the above). Any help with these questions is welcome. As you can tell, I have trouble starting on this. Thank you.
 
Last edited:
#2
Merlion - to get a "feel" for your data run a linear regression where the response variable is a function of the pretreatment variables i.e. y = a + b*x1 + c*x2 + d*x3 + ...

Next look at the absolute value of the t-statistics of the pretreatment variables - those greater than 2 have a significant influence. The coefficients of the pretreatment variables give the relative influence of each.