bugPSPP - Bugs: bug #35420, Regression output round off problem

 
 

bug #35420: Regression output round off problem

Submitted by:  None
Submitted on:  Tue 31 Jan 2012 10:31:51 PM UTC  
 
Category: Numerical ErrorsSeverity: 3 - Ordinary
Status: InvalidAssigned to: None
Open/Closed: ClosedRelease: None
Effort: 0.00

Add a New Comment(Rich Markup)
   

You are not logged in

Please log in, so followups can be emailed to you.

 

Fri 10 Feb 2012 01:43:06 PM UTC, comment #2:

Closing, since nothing needs to be done.

John Darrington <jmd>
Project Administrator
Wed 01 Feb 2012 07:48:38 AM UTC, comment #1:

Thanks very much for the report.

The "incorrect" value in physiology.sav is intentional! This file is used as an example in Section 5.2 (Data Screening and Transformation) of the User Manual. It demonstrates how PSPP can be used to catch typical data entry errors such as this.

Regarding the rounding issue, you can increase the reported number of decimal places by issuing the command "SET FORMAT F20.6." at the top of your file (this example will set the default output to a width of 20 and a precision of 6 decimal places). This is slightly different to how SPSS works.

John Darrington <jmd>
Project Administrator
Tue 31 Jan 2012 10:31:51 PM UTC, original submission:

I was experimenting with the correlation and regression features using the physiology.sav file from the example folder. I noticed that the residual for the variable "weight" for case 38 was very large. I then noticed that the weight for this case was apparently incorrect - it was negative.

I deleted the negative sign and ran the analysis again to see how much it improved. I was alarmed to see that the second run produced a multiple R of 1.00. Although the predictions improved, they are clearly not perfect.

I ran the analysis in SPSS, and found that the R was .998 - which would round up to 1.00 (and the R square of .996 would also round up to 1.00). Although 1.00 is technically correct, to two significant digits, I think that it's worth changing the output in cases like this - or printing a warning.

I have included the modified .sav file. RES1 and PRED1 were produced BEFORE I corrected the negative weight. RES2 and PRED2 were produced with the modified file. I have also included a pdf of the output file - the second run of the regression procedure was with the modified data file and shows the R of 1.00.

Anonymous

 

(Note: upload size limit is set to 16384 kB, after insertion of the required escape characters.)

Attach File(s):
   
   
Comment:
   

Attached Files
file #24946:  physiologyCorrected.sav added by None (3KiB - application/octet-stream)
file #24947:  physiologyCorrectedOutput.pdf added by None (35KiB - application/pdf)

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -unavailable- added by jmd (Posted a comment)
  •  

    Do you think this task is very important?
    If so, you can click here to add your encouragement to it.
    This task has 0 encouragements so far.

    Only logged-in users can vote.

     

    Please enter the title of George Orwell's famous dystopian book (it's a date):

     

     

    Follow 5 latest changes.

    Date Changed By Updated Field Previous Value => Replaced By
    Fri 10 Feb 2012 01:43:06 PM UTCjmdStatusNone=>Invalid
      Open/ClosedOpen=>Closed
    Wed 01 Feb 2012 07:48:38 AM UTCjmdSeverity5 - Average=>3 - Ordinary
    Tue 31 Jan 2012 10:31:51 PM UTCNoneAttached File-=>Added physiologyCorrected.sav, #24946
      Attached File-=>Added physiologyCorrectedOutput.pdf, #24947

    Back to the top


    Powered by Savane 3.1-cleanup1