Re: Deriving Formula from Ordinal Regression Results to Classify New Cases?

Posted by Vik Rubenfeld on Oct 24, 2012; 9:02am
URL: http://spssx-discussion.165.s1.nabble.com/Deciphering-from-t-test-the-mean-tp5715819p5715827.html

<base href="x-msg://240/">

Hi Ruben,

You were kind enough to bring this approach to my attention previously, and it was extremely useful in dividing the data into training and test data sets. Thanks very much!

In this case, the client has requested an Excel spreadsheet, into which they can enter the values for the independent variables. The spreadsheet would then calculate the predicted value of the DV.  So, I need to identify the correct formula.

Best,

-Vik

On Oct 23, 2012, at 11:59 PM, Ruben van den Berg wrote:

Dear Vik,

I guess if you figure out the structure of the formula (maybe "help" -> "algorithms" ?) we could insert the parameter estimates with Python (spssaux.GetValuesFromXMLWorkspace), thus completing your formula.

However, "if it ain't broken...". So what's the problem with the standard approach?

1) add the new cases to the old ones
2) compute a weight variable, 1 for old cases, close to 0 (e.g. 1e-12) for new ones
3) if necessary, compute a valid value on the dependent variable for new cases
4) weight cases by weight variable
5) run analysis and save predicted values

Including new cases in the analysis ensures they'll be given predicted values. However, the weighting procedure makes sure they won't have any effect on the parameter estimates (you'll see that these will be the same as for the old cases exclusively).

Best,

Ruben

> Date: Tue, 23 Oct 2012 20:30:29 -0700

> From: [hidden email]
> Subject: Deriving Formula from Ordinal Regression Results to Classify New Cases?
> To: [hidden email]
> 
> What is the correct method for deriving a formula from the results of an Ordinal Regression, that can be used to predict the value of the dependent variable for new cases?
> 
> Thanks very much in advance to all for any info.
> 
> Best,
> 
> 
> -Vik
> 
> =====================
> 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