How to test it?

Home > Statistics Every Writer Should Know > The Stats Board > Discusssion

How to test it?
Message posted by Julie (via 209.226.182.136) on November 24, 2001 at 2:54 PM (ET)

I was hoping you might know how to approach an analysis of a study i am proposing. I am studying what effects Test taking anxiety (TTA). My research question is, does procrastination(a behavioural trait) or specific traits chosen from the PRF (personality traits) contribute more to TTA. My measures are the PRF, the Procrastination Assessment Scale (PASS), and the Test Anxiety Inventory(TAI). The PRF is the only scale that will give me multiple numbers for each participant. The other two tests will provide me with one solid number each.

I figured that i would have to compare them by first doing a bivariate correlation on the PASS and TAI scores. Then i would find the r squared.

I could compare this r squared to the r squard that i obtain when i do a multiple correlation on the PRF with the TAI. Then i would know which is a better predictor, the PASS or the PRF.

I'm not sure if the above is the proper analysis.

But my main problem is then finding if there is incremental validity. I want to know if i say find that PRF is a better predictor of TTA, how do i find out if using procrastination as a predictor as well would be beneficial?

Thank you.

READERS RESPOND:
(In chronological order. Most recent at the bottom.)

Re: How to test it?
Message posted by Phil (via 165.247.234.102) on November 25, 2001 at 5:34 PM (ET)

You are referring to what some research methods texts call "Criterion Validity". Possibly multiple regression using the other measures would produce a better predictor. If you don't get a good "fit", try tranforming some of the variables and using stepwise or all possible regressions and try again. (It is not clear how many total measures you have). Also, don't know if analysis of covariance is called for using one of the measures as a co-variate.

Are you considering "Construct Validity" as well? Possible factor analysis could be used with all your measures to discover some "dimensions" or "constructs" in the measures that will give you new insight.

(hope that made some sense)

Your $5 contribution helps cover part the $500 annual cost of keeping this site online.