Discriminant Function Analysis

[rsling]

Situation:

If there is only 1 predictor (scoring of a particular test/questionnaire) and i would like to predict some group membership (more than 2 groups).

Would Discriminant Function Analysis be the right choice of statistical procedure to run?

Thanks!

[Bruce Weaver]

rsling wrote

Situation:

If there is only 1 predictor (scoring of a particular test/questionnaire) and i would like to predict some group membership (more than 2 groups).

Would Discriminant Function Analysis be the right choice of statistical procedure to run?

Thanks!

That is one option.  Logistic regression is another.  If there are more than 2 groups, then it would be multinomial logistic regression (NOMREG procedure).  One reason some folks prefer logistic regression is that there is no assumption of multivariate normality for the explanatory variables.  Explanatory variables can be categorical or quantitative.  And exponentiating the regression coefficients yields odds ratios.  There are chapters on both methods in David Garson's online stats notes, which you can find via Google.

[SR Millis-3]

I'd recommend logistic regression: it has fewer restrictive assumptions than DFA and is easier to interpret and implement the results.

Scott Millis


--- On Fri, 10/2/09, rsling <[hidden email]> wrote:

> From: rsling <[hidden email]>
> Subject: Discriminant Function Analysis
> To: [hidden email]
> Date: Friday, October 2, 2009, 2:56 AM
> Situation:
>
> If there is only 1 predictor (scoring of a particular
> test/questionnaire)
> and i would like to predict some group membership (more
> than 2 groups).
>
> Would Discriminant Function Analysis be the right choice of
> statistical
> procedure to run?
>
> Thanks!
> --
> View this message in context: http://www.nabble.com/Discriminant-Function-Analysis-tp25711210p25711210.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