Mixed Design ANOVA after multiple imputation

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

Mixed Design ANOVA after multiple imputation

LisaSnel
Dear everyone,

I have a question about performing a Mixed Design ANOVA in SPSS, after
multiple imputation. I have imputed my data, but when I perform the Mixed
Design ANOVA, I do not get a pooled result. The syntax I used is like this:

GLM Mg_preNTx Mg_D7 Mg_M1 Mg_M3 Mg_M6 Mg_M12 BY OBESITY
  /WSFACTOR=Magnesium 6 Polynomial
  /METHOD=SSTYPE(3)
  /PLOT=PROFILE(Magnesium*OBESITY)
  /CRITERIA=ALPHA(.05)
  /WSDESIGN=Magnesium
  /DESIGN=OBESITY.

Is it possible to generate a pooled p-value for this analysis? Because what
I want is one p-value (Sphericity assumed or with GreenhouseGeisser
correction) that says: are the Mg values over time different for patients
with and without obesity?


Thank you in advance!




--
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
Reply | Threaded
Open this post in threaded view
|

Re: Mixed Design ANOVA after multiple imputation

Joost van Ginkel
Dear Lisa,
 
You should read:
 
Van Ginkel, J.R. & Kroonenberg, P.M. (2014). Analysis of variance of multiply imputed data. Multivariate Behavioral Research, 49, 78-91. doi: 10.1080/00273171.2013.855890
 
The paper comes with an SPSS macro that is available from my personal page:
 
 
If you (and any other people reading this) need any help, please let me know.
 
Best regards,
 
Joost van Ginkel
 
-----Original Message-----
From: SPSSX(r) Discussion [[hidden email]] On Behalf Of LisaSnel
Sent: Tuesday, January 08, 2019 11:53 AM
To: [hidden email]
Subject: Mixed Design ANOVA after multiple imputation
 
Dear everyone,
 
I have a question about performing a Mixed Design ANOVA in SPSS, after
multiple imputation. I have imputed my data, but when I perform the Mixed
Design ANOVA, I do not get a pooled result. The syntax I used is like this:
 
GLM Mg_preNTx Mg_D7 Mg_M1 Mg_M3 Mg_M6 Mg_M12 BY OBESITY
  /WSFACTOR=Magnesium 6 Polynomial
  /METHOD=SSTYPE(3)
  /PLOT=PROFILE(Magnesium*OBESITY)
  /CRITERIA=ALPHA(.05)
  /WSDESIGN=Magnesium
  /DESIGN=OBESITY.
 
Is it possible to generate a pooled p-value for this analysis? Because what
I want is one p-value (Sphericity assumed or with GreenhouseGeisser
correction) that says: are the Mg values over time different for patients
with and without obesity?
 
 
Thank you in advance!
 
 
 
 
--
 
=====================
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: Mixed Design ANOVA after multiple imputation

Joost van Ginkel
Dear Eduard,

Firstly, the syntax doesn't provide variable names. Maybe that is something to include in future versions if I have the time. However, the F-values are displayed in the same order as the order in which they are displayed in the output of the (non-pooled) Mixed models analysis (additionally, you can also see it in the number of model degrees of freedom). To give a little bit more explanation, the COMBINED OVERALL TEST is the F-test which tests whether there are any significant differences, the COMBINED RESULTS provide the F-values of the separate effects (with the F-test for the intercept on top), and the Estimate, SE, t, df, and p are the results of the pooled regression coefficients of the model. In ANOVA you normally don't interpret these coefficients.
As for the correction for the violation of Sphericity assumption: In Mixed models there is an option for a Huyn-Feldt correction (under Random.., Covariance Type). If you select that option and you combine the results using my syntax, it will probably this correction into account in the pooling as well. I don't see any Greenhouse-Geisser correction option so that it probably not possible.

Best,

Joost

-----Original Message-----
From: SPSSX(r) Discussion [mailto:[hidden email]] On Behalf Of EduardNash
Sent: Wednesday, January 09, 2019 9:40 AM
To: [hidden email]
Subject: Re: Mixed Design ANOVA after multiple imputation

Dear Lisa and Joost,

Thank you for your question and answer. I am dealing with a similar issue: I
want to know if blood pressure over time is significantly different between
patients receiving treatment A and patients receiving treatment B. I also
performed a Mixed Design ANOVA and used the Sphericity assumption and the
Greenhouse-Geisser correction to get the p-value for significance. However,
I also needed to impute my data.

I struggled with similar issues as Lisa and tried the approach Joost wrote
down. But the output I get is strange:
Run MATRIX procedure:
 
COMBINED OVERALL TEST
      F-Value          df1          df2                   p
      69.5618      11.0000     716.7493        .0000
 
COMBINED RESULTS
      F-Value          df1               df2                p
   10448.4401       1.0000     156.9713        .0000
       3.8572       1.0000        156.5526        .0513
     106.4352       5.0000       754.7524        .0000
       2.2857       5.0000        727.0853        .0446
 
 
     Estimate           SE            t              df                  p
        .6997        .0068     102.2176     156.9713        .0000
       -.0135        .0069      -1.9640     156.5526        .0513
        .1752        .0088      19.9457     573.5050        .0000
        .0341        .0089       3.8328     461.3061        .0001
       -.1061        .0086     -12.3504     779.2266        .0000
       -.0737        .0087      -8.4362     625.9757        .0000
       -.0276        .0088      -3.1324     536.4563        .0018
        .0235        .0089       2.6547     490.5419        .0082
        .0099        .0093       1.0676     238.6150        .2868
       -.0078        .0086       -.9081     785.6911        .3641
       -.0181        .0087      -2.0703     634.4185        .0388
       -.0056        .0088       -.6349     511.7262        .5258
 
------ END MATRIX -----

In the combined results, I only see the F-statistic and not the variable
name. Besides that, I do not get a p-value for the Sphericity assumption
and/or Greenhous Geisser correction. And those two are the ones I am
interested in. So my question is: is there a way to get to those two
p-values?

I would very appreciate the help.

Best wishes,
Eduard Nash



--
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

=====================
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