Standardised/standardized residuals (Mixed)

classic Classic list List threaded Threaded
5 messages Options
Reply | Threaded
Open this post in threaded view
|

Standardised/standardized residuals (Mixed)

Steph Bullen
Hi all,

I have just constructed a growth curve model using SPSS (syntax below) and want to do some model checking with a standardised residual x predicted value scatter plot and histogram of standardised residuals. However, SPSS does not appear to recognising the ZRESID command in /SAVE. In addition, when this model is put in manually using the dialogue boxes there is no option to check "standardised residuals".

Has anybody encountered this before, and if so how did you circumvent it?

Thanks in advance.

Steph

Syntax:

 MIXED Weight BY Farm Group WITH Age Baseline_Weight Time
  /CRITERIA = CIN(95) MXITER(100) MXSTEP(5) SCORING(1)
  SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE)
  LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)
 /FIXED =Farm Group Baseline_Weight Age Time Farm*Time Farm*Age Farm*Baseline_Weight | SSTYPE(3)
  /METHOD = REML
  /PRINT =G SOLUTION TESTCOV
  /RANDOM =INTERCEPT Time | SUBJECT(ID) COVTYPE(UN)
 /SAVE =RESID PRED.
Reply | Threaded
Open this post in threaded view
|

Re: Standardised/standardized residuals (Mixed)

Maguin, Eugene
See page 1104 of the V22 reference manual (or its equivalent for other versions). However, standardized values are not available.

I do not know how to compute standardized values of either coefficients or residuals for mixed models. Perhaps a more knowledgeable person will respond.
Gene Maguin


-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of Steph Bullen
Sent: Monday, July 20, 2015 8:35 AM
To: [hidden email]
Subject: Standardised/standardized residuals (Mixed)

Hi all,

I have just constructed a growth curve model using SPSS (syntax below) and want to do some model checking with a standardised residual x predicted value scatter plot and histogram of standardised residuals. However, SPSS does not appear to recognising the ZRESID command in /SAVE. In addition, when this model is put in manually using the dialogue boxes there is no option to check "standardised residuals".

Has anybody encountered this before, and if so how did you circumvent it?

Thanks in advance.

Steph

Syntax:

 MIXED Weight BY Farm Group WITH Age Baseline_Weight Time
  /CRITERIA = CIN(95) MXITER(100) MXSTEP(5) SCORING(1)
  SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE)
  LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)  /FIXED =Farm Group Baseline_Weight Age Time Farm*Time Farm*Age Farm*Baseline_Weight | SSTYPE(3)
  /METHOD = REML
  /PRINT =G SOLUTION TESTCOV
  /RANDOM =INTERCEPT Time | SUBJECT(ID) COVTYPE(UN)  /SAVE =RESID PRED.



--
View this message in context: http://spssx-discussion.1045642.n5.nabble.com/Standardised-standardized-residuals-Mixed-tp5730188.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
Reply | Threaded
Open this post in threaded view
|

Re: Standardised/standardized residuals (Mixed)

Andy W
It seems one way would be to save RESID and SEPRED and then simply divide the former by the latter. This probably does not have a well defined distribution (like Studentized residuals in linear regression) but is probably reasonable for identifying outlying observations or obvious signs of model mis-fit. They should at least be symmetric I imagine.
Andy W
apwheele@gmail.com
http://andrewpwheeler.wordpress.com/
Reply | Threaded
Open this post in threaded view
|

Re: Standardised/standardized residuals (Mixed)

Bruce Weaver
Administrator
I was thinking along the same lines as Andy, FWIW.

Andy W wrote
It seems one way would be to save RESID and SEPRED and then simply divide the former by the latter. This probably does not have a well defined distribution (like Studentized residuals in linear regression) but is probably reasonable for identifying outlying observations or obvious signs of model mis-fit. They should at least be symmetric I imagine.
--
Bruce Weaver
bweaver@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.
Reply | Threaded
Open this post in threaded view
|

Re: Standardised/standardized residuals (Mixed)

Bruce Weaver
Administrator
This same basic question was posted to ResearchGate recently:

https://www.researchgate.net/post/SPSS_Mixed_Model_Linear_standardizing_residuals

While looking for info that might help the questioner, I found this old
thread, and was reminded that both Andy W and I had speculated that
RESID/SEPRED might give the desired standardized residual.  I decided to
test that idea by estimating a model via Stata's -mixed- command and saving
the standardized residuals via the -predict- post-estimation command.  Then
I estimated the same model via SPSS MIXED, and saved RESID and SEPRED.
Unfortunately, RESID/SEPRED does not match the standardized residuals from
Stata.  Details can be seen in the uploaded PDF file.*

MIXED_standardized_residuals.pdf
<http://spssx-discussion.1045642.n5.nabble.com/file/t7186/MIXED_standardized_residuals.pdf>  

Cheers,
Bruce

* I think that even if you do not use Nabble to read the list, you should
see a link to the uploaded file.  If not, you can visit the Nabble archive
and download it from there.  Here's the link to this thread:

http://spssx-discussion.1045642.n5.nabble.com/Standardised-standardized-residuals-Mixed-td5730188.html




Bruce Weaver wrote
> I was thinking along the same lines as Andy, FWIW.
> Andy W wrote
>> It seems one way would be to save RESID and SEPRED and then simply divide
>> the former by the latter. This probably does not have a well defined
>> distribution (like Studentized residuals in linear regression) but is
>> probably reasonable for identifying outlying observations or obvious
>> signs of model mis-fit. They should at least be symmetric I imagine.





-----
--
Bruce Weaver
[hidden email]
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.

--
Sent from: http://spssx-discussion.1045642.n5.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
--
Bruce Weaver
bweaver@lakeheadu.ca
http://sites.google.com/a/lakeheadu.ca/bweaver/

"When all else fails, RTFM."

NOTE: My Hotmail account is not monitored regularly.
To send me an e-mail, please use the address shown above.