Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
| From | "Feiveson, Alan H. (JSC-SK311)" <alan.h.feiveson@nasa.gov> |
| To | "statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu> |
| Subject | st: -anova- command fails with mildly complexx model |
| Date | Thu, 15 Apr 2010 16:15:04 -0500 |
Hello - I have Stata 11 IC running on Windows XP. When I try to run -anova- on the data set I have listed at the end of this message, bad things happen. Please see below. I would welcome any suggestions or explanations.
Al Feiveson
My factors are:
group (2 levels)
treat(2 levels)
isub (30 levels - nested within group # treat)
time (4 levels, repeated for each isub)
y is the dependent variable
First I run this in "non-repeated" syntax without the group#treat interaction. It works.
. use anovatest_041410
. anova y treat group / isub|treat#group time time#(group##treat)
Number of obs = 120 R-squared = 0.9518
Root MSE = .504486 Adj R-squared = 0.9265
Source | Partial SS df MS F Prob > F
-----------------+----------------------------------------------------
Model | 392.423133 41 9.57129592 37.61 0.0000
|
treat | 145.674315 1 145.674315 30.65 0.0000
group | 100.00584 1 100.00584 21.04 0.0001
isub|treat#group | 128.34615 27 4.75356109
-----------------+----------------------------------------------------
time | .899050531 3 .29968351 1.18 0.3238
time#group | .008263434 3 .002754478 0.01 0.9984
time#treat | 1.06352637 3 .354508791 1.39 0.2512
time#group#treat | .768763895 3 .256254632 1.01 0.3943
|
Residual | 19.8514669 78 .254505985
-----------------+----------------------------------------------------
Total | 412.2746 119 3.46449243
Now I try to add the group#treat interaction term:
anova y treat group treat#group / isub|treat#group time time#(group##treat)
[stuck in infinite loop]
close out Stata, bring it up and try again with the interaction term:
use anovatest_041410
anova y treat group treat#group / isub|treat#group time time#(group##treat)
[crashed] The instruction at "x7c911129"refrenced memory at "0x031faad5". The memory could not be "read".
close out Stata, bring it up and try again without the "time" part of the model. It works.
use anovatest_041410
. anova y treat group treat#group / isub|treat#group
Number of obs = 120 R-squared = 0.9451
Root MSE = .501472 Adj R-squared = 0.9274
Source | Partial SS df MS F Prob > F
-----------------+----------------------------------------------------
Model | 389.641905 29 13.4359278 53.43 0.0000
|
treat | 148.76034 1 148.76034 31.22 0.0000
group | 100.899184 1 100.899184 21.18 0.0001
treat#group | 20.7962645 1 20.7962645 4.36 0.0466
isub|treat#group | 123.880297 26 4.76462679
-----------------+----------------------------------------------------
|
Residual | 22.6326945 90 .251474384
-----------------+----------------------------------------------------
Total | 412.2746 119 3.46449243
Now add the "time" part back in:
anova y treat group treat#group / isub|treat#group time time#(group##treat)
[crashed] The instruction at "0x00727e9e"refrenced memory at "0x00000000". The memory could not be "read".
Bring up Stata and try again. This time it ran, but the factors are all messed up. Note that there is no treat#group factor in the anova table, and that "time" appears in the upper part where it doesn't belong. Also no treat#group interaction.
use anovatest_041410
. anova y treat##group / isub|treat#group time time#(group##treat)
Number of obs = 120 R-squared = 0.9518
Root MSE = .504486 Adj R-squared = 0.9265
Source | Partial SS df MS F Prob > F
-----------------+----------------------------------------------------
Model | 392.423133 41 9.57129592 37.61 0.0000
|
treat | 145.674315 1 145.674315 30.65 0.0000
group | 100.00584 1 100.00584 21.04 0.0001
time#group | .008263434 3 .002754478 0.00 1.0000
isub|treat#group | 128.34615 27 4.75356109
-----------------+----------------------------------------------------
time | .899050531 3 .29968351 1.18 0.3238
time#treat | 1.06352637 3 .354508791 1.39 0.2512
time#treat#group | .768763895 3 .256254632 1.01 0.3943
|
Residual | 19.8514669 78 .254505985
-----------------+----------------------------------------------------
Total | 412.2746 119 3.46449243
Now try it with the -repeated- option. First without the treat#group interaction. It works.
. anova y treat group / isub|treat#group time time#(group##treat) ,repeated(time)
Number of obs = 120 R-squared = 0.9518
Root MSE = .504486 Adj R-squared = 0.9265
Source | Partial SS df MS F Prob > F
-----------------+----------------------------------------------------
Model | 392.423133 41 9.57129592 37.61 0.0000
|
treat | 145.674315 1 145.674315 30.65 0.0000
group | 100.00584 1 100.00584 21.04 0.0001
isub|treat#group | 128.34615 27 4.75356109
-----------------+----------------------------------------------------
time | .899050531 3 .29968351 1.18 0.3238
time#group | .008263434 3 .002754478 0.01 0.9984
time#treat | 1.06352637 3 .354508791 1.39 0.2512
time#group#treat | .768763895 3 .256254632 1.01 0.3943
|
Residual | 19.8514669 78 .254505985
-----------------+----------------------------------------------------
Total | 412.2746 119 3.46449243
Between-subjects error term: isub|treat#group
Levels: 30 (27 df)
Lowest b.s.e. variable: isub
Covariance pooled over: treat#group (for repeated variable)
Repeated variable: time
Huynh-Feldt epsilon = 1.0920
*Huynh-Feldt epsilon reset to 1.0000
Greenhouse-Geisser epsilon = 0.9061
Box's conservative epsilon = 0.3333
------------ Prob > F ------------
Source | df F Regular H-F G-G Box
-----------------+----------------------------------------------------
time | 3 1.18 0.3238 0.3238 0.3225 0.2878
time#group | 3 0.01 0.9984 0.9984 0.9973 0.9179
time#treat | 3 1.39 0.2512 0.2512 0.2535 0.2486
time#group#treat | 3 1.01 0.3943 0.3943 0.3894 0.3249
Residual | 78
----------------------------------------------------------------------
Now try again with the treat#group interaction added. This time it thinks "time" is the error term for the first part and various other terms are missing.
. anova y treat##group / isub|treat#group time time#(group##treat) ,repeated(time)
Number of obs = 120 R-squared = 0.9518
Root MSE = .504486 Adj R-squared = 0.9265
Source | Partial SS df MS F Prob > F
-----------------+----------------------------------------------------
Model | 392.423133 41 9.57129592 37.61 0.0000
|
treat | 145.674315 1 145.674315 486.09 0.0002
group | 100.00584 1 100.00584 333.70 0.0004
isub|treat#group | 130.023122 27 4.81567118 16.07 0.0207
time | .899050531 3 .29968351
-----------------+----------------------------------------------------
time#treat | 1.06352637 3 .354508791 1.39 0.2512
time#treat#group | .78408852 6 .13068142 0.51 0.7964
|
Residual | 19.8514669 78 .254505985
-----------------+----------------------------------------------------
Total | 412.2746 119 3.46449243
Between-subjects error term: isub|treat#group
Levels: 30 (27 df)
Lowest b.s.e. variable: isub
Covariance pooled over: treat#group (for repeated variable)
Repeated variable: time
Huynh-Feldt epsilon = 1.0920
*Huynh-Feldt epsilon reset to 1.0000
Greenhouse-Geisser epsilon = 0.9061
Box's conservative epsilon = 0.3333
------------ Prob > F ------------
Source | df F Regular H-F G-G Box
-----------------+----------------------------------------------------
time#treat | 3 1.39 0.2512 0.2512 0.2535 0.2486
time#treat#group | 6 0.51 0.7964 0.7964 0.7796 0.6044
Residual | 78
----------------------------------------------------------------------
Data:
isub treat group time y
1 0 0 1 3.128386
1 0 0 2 3.121638
1 0 0 3 2.875379
1 0 0 4 2.561247
2 0 0 1 1.007731
2 0 0 2 1.717744
2 0 0 3 2.074634
2 0 0 4 2.258288
3 0 0 1 3.533892
3 0 0 2 3.605727
3 0 0 3 2.435609
3 0 0 4 2.814991
4 0 0 1 .165792
4 0 0 2 1.45446
4 0 0 3 .0822014
4 0 0 4 1.636633
5 0 0 1 1.549338
5 0 0 2 1.182068
5 0 0 3 1.227366
5 0 0 4 .7778459
6 0 0 1 1.173828
6 0 0 2 .8148147
6 0 0 3 1.368575
6 0 0 4 -.0073038
7 0 0 1 3.100238
7 0 0 2 2.073025
7 0 0 3 2.476525
7 0 0 4 2.520768
8 0 0 1 2.807955
8 0 0 2 2.805253
8 0 0 3 2.153584
8 0 0 4 2.5705
9 0 0 1 2.704864
9 0 0 2 2.654852
9 0 0 3 2.746811
9 0 0 4 2.94425
10 0 0 1 1.6997
10 0 0 2 1.810207
10 0 0 3 1.55903
10 0 0 4 1.669906
11 1 0 1 1.868059
11 1 0 2 3.322237
11 1 0 3 3.7292
11 1 0 4 3.670806
12 1 0 1 2.434988
12 1 0 2 3.323258
12 1 0 3 3.372895
12 1 0 4 3.338208
13 1 0 1 1.22023
13 1 0 2 .7378999
13 1 0 3 .5126179
13 1 0 4 1.54094
14 1 0 1 4.453463
14 1 0 2 4.22743
14 1 0 3 5.309576
14 1 0 4 5.050639
15 1 0 1 4.416439
15 1 0 2 3.452432
15 1 0 3 4.392283
15 1 0 4 4.605953
16 1 0 1 5.449149
16 1 0 2 4.524527
16 1 0 3 4.718941
16 1 0 4 4.981455
17 1 0 1 2.48961
17 1 0 2 3.723629
17 1 0 3 2.284816
17 1 0 4 3.453557
18 1 0 1 3.641483
18 1 0 2 3.406856
18 1 0 3 3.208734
18 1 0 4 3.80458
19 0 1 1 3.37035
19 0 1 2 3.97767
19 0 1 3 3.014043
19 0 1 4 3.020998
20 0 1 1 1.288446
20 0 1 2 1.037267
20 0 1 3 1.188551
20 0 1 4 1.292643
21 0 1 1 3.397913
21 0 1 2 3.811652
21 0 1 3 3.757371
21 0 1 4 3.876024
22 0 1 1 3.147866
22 0 1 2 3.404653
22 0 1 3 3.840606
22 0 1 4 3.757992
23 0 1 1 4.017531
23 0 1 2 4.78287
23 0 1 3 4.696186
23 0 1 4 3.947599
24 0 1 1 2.150811
24 0 1 2 1.366459
24 0 1 3 2.619952
24 0 1 4 2.440276
25 1 1 1 3.84514
25 1 1 2 4.948975
25 1 1 3 4.019058
25 1 1 4 5.542306
26 1 1 1 7.858982
26 1 1 2 7.427705
26 1 1 3 7.378307
26 1 1 4 8.286385
27 1 1 1 5.736018
27 1 1 2 7.030629
27 1 1 3 6.816757
27 1 1 4 7.087567
28 1 1 1 5.363531
28 1 1 2 4.517203
28 1 1 3 4.877666
28 1 1 4 5.66123
29 1 1 1 6.778078
29 1 1 2 6.779854
29 1 1 3 6.729872
29 1 1 4 5.738735
30 1 1 1 7.368432
30 1 1 2 6.341629
30 1 1 3 5.96716
30 1 1 4 6.480004
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/