Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: st: RE: Use of data originating from SPSS

From   Nick Cox <>
To   "" <>
Subject   RE: st: RE: Use of data originating from SPSS
Date   Thu, 2 Feb 2012 21:55:58 +0000

-mfp- wants numeric variables. 

I see no reason why value labels would inhibit anything whatsoever that it does. 

The issue is perhaps just that you have too many missings. 


Margaret MacDougall [] 

Thanks for the valued prompt session, Nick. I don't see any sign of my 
variables appearing as string. They are all being recognized in Stata as 
being of type double and format %10.0g.

Any further thoughts?  I am wondering if the mfp procedure mentioned in 
my first email requires all variable values or at least those of the 
dependent variable to appear as numbers (without the value labels 
showing in the case of categorical variables). I could investigate this 
possibility but further tried and tested ideas to redeem me form the 
trial and error approach would be welcome.

On 01/02/2012 17:31, Nick Cox wrote:

> I don't use SPSS but you should check which variables are regarded as string by Stata. -describe- or -varm- will tell you this. You may need to -encode- or -destring- some variables.

Margaret MacDougall

> I have recently migrated from SPSS to Stata to allow me to implement
> fractional polynomial techniques, including the mfp procure, for a
> logistic regression analysis. I have been able to save the original SPSS
> file in Stata format. The variables in the SPSS file were of numeric
> "Type" (in the sense intended by the SPSS Variable View facility) with
> text labels linked to the numerical codes using the SPSS Value Labels
> function. On opening up the Stata version of this file, the text labels
> appear in the individual columns for the dependent and independent
> variables. When I run the code for identifying the Stata file and
> running the mfp procedure, the error code r(2000) is generated, pointed
> to an absence of observations.
> Is it possible that I require to reformat the data in some way for the
> procedure to run successfully?

*   For searches and help try:

© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index