sas >> How do you calculate Eta Squared using Proc Mixed?

by hodab » Thu, 17 Feb 2005 01:56:34 GMT

Hi all:

I have computed a repeated measures multilevel models analysis in SAS
using Proc mixed on some EMA data.
It was a pretty simple analysis looking at the degree of association
between two variables: mood and pain with time as the repeated measure.
Well, the editor of the journal I am sending the article to would like
me to report effect sizes for the analysis. I know how to get SAS to
print out a 95% confidence interval in Proc Mixed, but I cannot get it
to print out the Type III Sums of Squares that I would need to
calculate Eta squared. My problem, is that Proc Mixed uses MLE and
does not calculate Type III SS. My understanding is that you can get
it to calculate them if you put in a "Method = Type 3" statement, but,
because I have a repeated models analysis and am already using the
"REPEATED" statement, I cannot do both. Do you know any way for me to
get SAS to calculate Eta Squared? If not, do you think that reporting
the confidence interval along with the p value would be sufficient?
Thanks!



sas >> How do you calculate Eta Squared using Proc Mixed?

by stringplayer_2 » Thu, 17 Feb 2005 03:36:05 GMT






What covariance structure is fit by your REPEATED statement. If
you have a TYPE=CS covariance structure, then you can use an
equivalent RANDOM statement. The following two models produce
identical results:

proc mixed data=mydata;
class id time;
model pain = mood / s;
repeated time / subject=id type=cs;
run;

proc mixed data=mydata;
class id;
model pain = mood / s;
random id;
run;

You could specify method=type3 to the latter model to get the
variance partitioning that you need to construct eta.

Alternatively, you can fit the model with and without mood.
Each produces an estimate of the residual variance. If we have

V1=residual variance from model without mood
V2=residual variance from model with mood

then an estimate of eta would be 1 - (V2/V1).

HTH,

Dale


=====
---------------------------------------
Dale McLerran
Fred Hutchinson Cancer Research Center
mailto: XXXX@XXXXX.COM
Ph: (206) 667-2926
Fax: (206) 667-5977
---------------------------------------




__________________________________
Do you Yahoo!?
Yahoo! Mail - You care about security. So do we.
http://promotions.yahoo.com/new_mail



sas >> How do you calculate Eta Squared using Proc Mixed?

by hodab » Thu, 17 Feb 2005 06:23:37 GMT

Dale -

Thanks so much for your help! It worked! :-)

take care,

hoda

___________________
Hoda Badr, Ph.D.
The University of Texas M. D. Anderson Cancer Center
email: XXXX@XXXXX.COM
____________________



Similar Threads

1. how to calculate sum of squares for treatments when using proc mixed

2. Effect size (eta squared, omega squared) in PROC MIXED?

Hello all,

Many journals request some estimate of effect size be included with 
statistical analyses.  However, as I understand it, there are some 
problems with such estimates in repeated measures designs.  I did find 
some references to a generalized eta or omega squared statistic devised 
by Olejnik & Algina (Psych Meth 2003, 8(4):434-447), which can be 
applied to repeated measures designs.  Has anyone implemented these as 
SAS macros for use with PROC MIXED?  (I know how to get the Type III 
Sums of Squares by forcing the method to type3, but that's about as far 
as I've gotten.)

Also, has anyone on the group tried these generalized statistics with 
designs more complex than in the manuscript? -- I'm trying to do eta 
squared with a model they would probably consider a "mixed design", and 
code as ABCDe (A-D all manipulated factors, with subjects nested in A, 
and fully repeated in factors B, C, and D, with subject as a "measured" 
factor e), but the effect sizes seem far too low, or I'm doing something 
wrong (the more likely situation).  (Would omega work despite the large 
number of factors in the model?)

Thanks!

3. how to calculate sum of squares for treatments when using

4. score chi-square & wald chi-square in logistic regression using sas

What is the difference between score chi-square & wald chi-square in
logistic regression?

5. r-squared in proc mixed

6. proc mixed / Latin squares+ repeated measures

Hi listers,

I would like to know if a Latin square design with repeated measures
within each cell can be set up with PROC MIXED. The Latin square has as
blocking effects, period (row) and animal (column) (both random
effects), the response of each animal at each treatment is measured four
times on time.
What term can be used as error for the between subjects design?
If period was fixed, then could I use the interaction treatment*animal
as error for the Latin square?
Thanks a lot !

7. R square in proc mixed?

8. Using proc freq : add column after doing freq

Hi Every one,


i am doing following

proc freq data= ****;
tables a*b*e/ list out= abc;
run;


i have normal abnormal and not reported in B,
 were as i have only normal and abnormal  in E, i dont have not
reported.
here a is 1;


this is how my output  looks like




                                                                  |
Normal  |Abnormal|  Total
                                                --------------+--------
+--------+
                                                Normal
9          0              9
                                             ---------------+--------
+--------+
                                             Abnormal       |
0        3               3
                                              --------------+--------
+--------+
                                             Not Reported *
1         0              1
                                             ---------------+--------
+--------+
                                             Total
10        31            13
      in data set abc:

                                    obs  A
B                          E      COUNT    PERCENT

                                     1      1    Normal
Normal          99        75.5725
                                     2      1    Abnormal
Abnormal      31        23.6641
                                     3      1    Not Reported *
Normal         1          0.7634




How do i add not reported in E so my o/p looks like


                                                                  |
Normal  |Abnormal|  not reported  Total
                                                --------------+--------
+--------+
                                                Normal
9          0                0               9
                                             ---------------+--------
+--------+
                                             Abnormal       |
0        3                0                3
                                              --------------+--------
+--------+
                                             Not Reported *
1         0               0                  1
                                             ---------------+--------
+--------+
                                             Total
10        31               0            13



Thanks,
Raghu