bugPSPP - Bugs: bug #51373, Problems with REGRESSION...

 
 

bug #51373: Problems with REGRESSION (Incorrect standard error?)

Submitted by:  None
Submitted on:  Mon 03 Jul 2017 05:33:12 PM UTC  
Votes:  100  
 
Category:  Numerical Errors Severity:  7 - Major
Status:  Fixed Assigned to:  None
Open/Closed:  Closed Release:  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.

 

Sat 29 Jul 2017 05:40:56 AM UTC, comment #5: 

I fixed this issue in commit 8961943f7492c63ba29cbb61faf698b59aca4ff2

It seems to be the result of an optimisation. 
I don't see that this optimisation actually brings us much.  So for now I have merely disabled it.

John Darrington <jmd>
Project Administrator
Fri 28 Jul 2017 03:01:58 PM UTC, comment #4: 

I can see the problem too.
I will try to find a fix shortly.

John Darrington <jmd>
Project Administrator
Wed 26 Jul 2017 06:35:36 AM UTC, comment #3: 

I could not see my attached data file here and am not sure why.

Yang <statlover>
Wed 26 Jul 2017 05:30:46 AM UTC, comment #2: 

I got the 2014 GSS data online (see the attached file) and tried to conduct the regression analyses. I also found the same issue: the standard errors and t values are different when the order of independent variables (pasei10 & masei10) changes. Can someone help fix this problem? Thank you!

Yang <statlover>
Sat 08 Jul 2017 04:44:41 AM UTC, comment #1: 

Can you please post the dataset that you used to do this.

Perhaps you can also state which version you are using and which operating system.

John Darrington <jmd>
Project Administrator
Mon 03 Jul 2017 05:33:12 PM UTC, original submission:  

I've searched through the list of bugs briefly, but didn't see anything listed...

I've been working with PSPP as an alternative for SPSS in my undergrad stats class. I noticed some irregularities with the linear regression module.

It appears that PSPP will produce different standard errors for the constant depending on the order in which the variables are entered into the model. This in turn produces different t-values and significance levels.

I've attached an output files using the 2014 GSS survey showing how different coefficient are produced depending on the order in which they are entered. I've also attached output from SPSS that can be used as a comparison.

Anonymous

 

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

Attach Files:
   
   
Comment:
   

Attached Files
file #41109:  SPSS Regression Output.pdf added by None (18KiB - application/pdf)
file #41110:  PSPP Regression Output.pdf added by None (38KiB - application/pdf)

 

Depends on the following items: None found

Items that depend on this one: None found

 

Carbon-Copy List
  • -email is unavailable- added by statlover (Posted a comment)
  • -email is unavailable- added by statlover (Voted in favor of this item)
  • -email is unavailable- added by jmd (Posted a comment)
  •  

    Do you think this task is very important?
    If so, you can add your encouragement to it.
    This task has 100 encouragements so far.

    Only logged-in users can vote.

     

     

     

    Follow 7 latest changes.

    Date Changed by Updated Field Previous Value => Replaced by
    2017-08-01 jmd StatusConfirmed => Fixed
        Open/ClosedOpen => Closed
    2017-07-28 jmd Severity5 - Average => 7 - Major
        StatusNone => Confirmed
    2017-07-26 statlover Carbon-Copy- => Added statlover
    2017-07-03 None Attached File- => Added SPSS Regression Output.pdf, #41109
        Attached File- => Added PSPP Regression Output.pdf, #41110

    Back to the top


    Powered by Savane 3.6