Re: What Statistical Analysis to use?

Posted by Art Kendall on Jun 09, 2012; 1:34pm
URL: http://spssx-discussion.165.s1.nabble.com/What-Statistical-Analysis-to-use-tp5713606p5713609.html

I assume that the scales are established measures of your variables

It is usually unnecessary to treat a summative scale as purely ordinal. Likert scales are usually considered _not severely discrepant_ from interval level.
Even single items are frequently treated as _not severely discrepant_ from interval level, but I presume you'll be using the summative scores.  I advise that you stick to more common approaches.

Self esteem could well be more of a trait variable than a state variable.  Don't be surprised if the pre-test accounts for a lot of the variance in the post test and does not leave much for other variables to account for.

I doubt that you have the power to detect interaction effects. I wouldn't be surprised if some of the contingencies added to the prediction, but why do you expect a interactions?

Also you are doing a lot of tests on a small amount of data, take a lot of salt with your results.

Center your CSW and achievement independent variables so you can create variables representing an interaction effect. (I.e., subtract the mean from the scores.)  Since the scaling is arbitrary this should not hinder interpretation.
Create 7 new variables representing interaction terms by multiplying the centered achievement variable by one of the centered CSW variables.

in REGRESSION use a stepped (aka a hierarchical, NOT a stepwise) approach. Much of what you would be looking for would be in the variables not yet in the equation.  Be sure to output the change in R**2.

For the first hypothesis the variables would be post test pre-test and achievement.
the post test self esteem would be the dependent variable.
enter the pre-test
read the output so see if achievement would significantly improve the prediction.  Remember that you are only finding out whether the improvement would be statistically distinguishable from no improvement.  This says nothing about whether the improvement is meaningful. 

For the other hypotheses the variables would be post test, pre-test, centered achievement, centered CSW variable, and the interaction term.
The dependent variable is the post-test self esteem
enter pre-test, centered achievement,centered CSW variable.
read the output so see if the interaction term  would significantly improve the prediction.  Remember that you are only finding out whether the improvement would be statistically distinguishable from no improvement.  This says nothing about whether the improvement is meaningful. 

HTH

On 6/9/2012 7:17 AM, wordcount wrote:

The CSW is a 7-point (ranging from Strongly Disagree to Strongly Agree)
Lickert scale with 7 subscales (Family Support; Competition; Appearance;
God's Love; Academic Competence; Virtue; Approval from Others) as given in
the 7 hypotheses I've formed. This scale has 35 total items and each
subscale has 5.

Self-esteem was measured by the Rosenberg's Self-Esteem Scale, which is a
4-point (ranging from Strongly Agree to Strongly Disagree) Lickert scale of
10 items. This is a dependent variable only, while Academic Performance as
measured by the marks obtained by the students on a 5-item test, each
carrying 5 marks (total 25 marks), and the contingencies of self-worth are
independent variables.

The sample size is 37.

--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/What-Statistical-Analysis-to-use-tp5713606p5713608.html
Sent from the SPSSX Discussion mailing list archive at Nabble.com.

=====================
To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD

===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD