I am testing a hypothesis in which my scale DV (PSS) is predicted by an interaction between an Ordinal DV and a scale IV.
The ordinal IV is TF (Total Freq)
The scale IV is and PQM (Practice Quality).
TF is the sum of two ordinal variables:FF(Formal Freq) and IF(Informal Freq)
Each is measured as follows:
A) never/almost never, B) 1/week, C) 2-3/week, D)Almost Everyday, E) Everyday, F) Two or more times/day, More than three times/day]
PQM Is a validated scale made up of the means of two sub scales comprising 3 questions which are each being measured from 0% to 100%.
My hypothesis is that high scores on TF (a composite of FF and IF) predict lower PSS (DV) scores only when PQM is also high.
Given that FF and IF are measuring frequency using unequal intervals(i.e. jump from 2 times a week to 'almost everyday'), what are my options for testing my hypothesis?
In terms of combining them: I figured I would centre each variable by subtracting the mean from each score and then add FF(centred) and IF(centred) to get TF
I currently think I have these options:
1) Do the regression assuming my ordinal DV's are scale
2) Dummy code my ordinal DVs and do the regression (Low vs High; coding them 0 and 1)
3) Do a one way ANOVA after reducing all variables to categories of Low/High
4) Do LOGARITHMIC TRANSFORMATIONS on ordinal variables and reduce scale variables to Low/High
In any of these cases I am not sure if I should enter TF alone or instead enter FF and IF together. I am also wondering the same thing about PQM which is made up of two opposing subclass (positive measure and negative measure).
The rationale for entering FF and IF separately is that the literature shows that only one predicts my DV and not the other.
I have a small sample of only about N=34. If list-wise deletion is used the total N is only around 25. **
I have two ordinal IVs (unequal intervals within each) , one scale IV and a scale DV and am wondering a) what test to use to test for an interaction between my DVs b) what transformations I need to perform and c) how to best calculate and enter composite measures into the statistical test.
Thank you if you got this far! I look forward to hearing your responses.