Which statistical test to use?

#1
Hi,

I'm struggling with the following;
I have two dependent variables (A and B) which are continous and highly correlated with each other (Pearson 0.9). I want to find out if A>B, A=B or A<B. The most logical test would be to perform a paired samples T test in SPSS, but I also want to correct for possible confounders (age, sex, etc.). Which test should I use? I'm thinking of a repeated measures MANCOVA, whereby I enter A and B as repeated measures, but I'm not getting output about the relationship between A and B. Does anyone have a suggestion on how to solve this?

Many thanks,

Frank
 

Karabiner

TS Contributor
#2
So the title of this thread is not only very vague, but also somewhat misleading, since you've already chosen your analysis? Are you asking a software(SPSS-)specific question, then? In that case it would be helpful if you used the SPSS-subforum and if you explained what you actually did when you "entered A and B as repeated measures" and what you mean by "I'm not getting output about the relationship between A and B". Some ideas about program handling you maybe can get from http://www.statisticshell.com/docs/repeatedmeasures.pdf , but there might be even better sources on the net.

With kind regards

K.
 
#3
....but I'm not getting output about the relationship between A and B. Does anyone have a suggestion on how to solve this?
Yes, show the print out! Maybe the result is written as a parameter estimate with a name that you are not familiar with. :)

But does it matter to include factors like age and gender? Doesn't the pairwise t-test "act as its own control", so that gender and age effects disappear when the difference is taken? Or is it just if there is an interaction effect between the treatment (A vs B) and age?