SAS output for exemplar 6 from scores

This file was produced using the SAS html output with the minimal style.

You can also view the program that created this output.

Some comments have been added to the output in blue and preceeded by **** 
Links in this page

First analyses of just the 6 prevalence measures
Analysis with repeated measures
Analysis with IVEWARE macro
Postimputation procedures for IVEWARE results
Analysis with PROC MI
Postimputation procedures for PROC MI results
Then with all scores and other varaibles 
Analysis with IVEWARE macro
Some data checks for IVEWARE results
Postimputation procedures for IVEWARE results
Analysis with PROC MI
Postimputation procedures for PROC MI results



**** This extract from the output gives the fitted means for the prevalence over time
back to top uparrow





 Least Squares Means




 Effect
 time
 Estimate
 Standard Error
 DF
 t Value
 Pr > |t|






 time
 1
 0.7323
 0.007032
 4327
 104.13
 <.0001




 time
 2
 0.7293
 0.006950
 4327
 104.94
 <.0001




 time
 3
 0.7843
 0.006399
 4327
 122.56
 <.0001




 time
 4
 0.7495
 0.006778
 4327
 110.57
 <.0001




 time
 5
 0.7209
 0.007215
 4327
 99.91
 <.0001




 time
 6
 0.6068
 0.008203
 4327
 73.98
 <.0001







back to top uparrow
**** Now some selected output from the IVEWARE results
First one of the equations. Details of every fit plus variance covariance 
matrices of parameters can be output.

Impute dprev6                                                                                                                        
                                                                                                                                     
Code: 0                                                                                                                              
                                                                                                                                     
Unperturbed and perturbed coefficients                                                                                               
                                                                                                                                     
Intercept       -1.14668586      -1.227304752                                                                                        
   GENDER     -0.2619741459     -0.3521203356                                                                                        
   dprev3        0.60660311      0.7099782128                                                                                        
   dprev4       0.637764151      0.7070447766                                                                                        
   dprev2      0.1090258038       0.158246624                                                                                        
   dprev1      0.2083380484      0.2469818953                                                                                        
   dprev5       1.562734838       1.617341827  

   **** At the end observed and imputed values are given. These are prevalence at sweep 5 and 6.
   YOu can see that the imputed data have somewhat higher prevalences

Variable dprev5                                                                                                                      
                  Observed           Imputed          Combined                                                                       
    Code       Freq    Per       Freq    Per       Freq    Per                                                                       
       0       1086  28.94        135  23.44       1221  28.21                                                                       
       1       2666  71.06        441  76.56       3107  71.79                                                                       
   Total       3752 100.00        576 100.00       4328 100.00                                                                       
                                                                                                                                     
Variable dprev6                                                                                                                      
                  Observed           Imputed          Combined                                                                       
    Code       Freq    Per       Freq    Per       Freq    Per                                                                       
       0       1391  40.98        325  34.80       1716  39.65                                                                       
       1       2003  59.02        609  65.20       2612  60.35




back to top uparrow
   **** Now extract from PROC MIANALYZE that gives mean of each score. This for boys only. From observed data only would be
         0.7895 0.7768 0.81100 0.7812 0.7266 0.6472, so not much affected by imputation.





 Multiple Imputation Parameter Estimates




 Parameter
 Estimate
 Std Error
 95% Confidence Limits
 DF
 Minimum
 Maximum
 Theta0
 t for H0:
Parameter=Theta0		
 Pr > |t|






 dprev1
 0.790544
 0.009190
 0.772452
 0.808635
 277.44
 0.787117
 0.793056
 0
 86.02
 <.0001




 dprev2
 0.780950
 0.009051
 0.763195
 0.798705
 1440.1
 0.778438
 0.782092
 0
 86.28
 <.0001




 dprev3
 0.813842
 0.008533
 0.797102
 0.830582
 1251.1
 0.811786
 0.815441
 0
 95.38
 <.0001




 dprev4
 0.784947
 0.009060
 0.767164
 0.802731
 830.06
 0.782092
 0.786661
 0
 86.64
 <.0001




 dprev5
 0.739493
 0.009759
 0.720321
 0.758664
 525.56
 0.737323
 0.742805
 0
 75.78
 <.0001




 dprev6
 0.649726
 0.011553
 0.626630
 0.672822
 61.586
 0.645500
 0.655550
 0
 56.24
 <.0001








back to top uparrow

   **** PROC MI produces various outputs including means of imputed variables,
   but this is not relevant here because this is on the unrounded data.
   Instead we show the range of imputed values obtained before rounding to 0 or 1.

   They all look OK.
Variable Mean Std Dev Minimum Maximum
_Imputation_
CASEID
GENDER
dprev1
dprev2
dprev3
dprev4
dprev5
dprev6		 5.5000000
3303.21
1.4942237
0.7321019
0.7301960
0.7844779
0.7501388
0.7205060
0.6061606		 2.8723145
1325.65
0.4999724
0.4439232
0.4455094
0.4132068
0.4348281
0.4516458
0.4889685		 1.0000000
1000.00
1.0000000
-0.8384172
-0.8728717
-0.6935567
-0.9319122
-0.8990870
-1.1732459		 10.0000000
5596.00
2.0000000
2.3938461
2.2746336
1.9690113
2.1037230
2.3992527
2.3684941		 

back to top uparrow
   **** Now same table after rounding to 0 1 gives the estimated prevalences. In fact 
  the means are not much affaected but obviously range is different.
Variable N Mean Std Dev Minimum Maximum
dprev1
dprev2
dprev3
dprev4
dprev5
dprev6		 43280
43280
43280
43280
43280
43280		 0.7295518
0.7277033
0.7828096
0.7469039
0.7164510
0.5994224		 0.4441965
0.4451469
0.4123380
0.4347906
0.4507257
0.4900212		 0
0
0
0
0
0		 1.0000000
1.0000000
1.0000000
1.0000000
1.0000000
1.0000000		 

back to top uparrow
   **** Imputation of the larger set of variables produces lots
of diagnostics too, all of which seem OK. Different variables 
had to be defined as the correct variable type. Simple procedures
like checking the range of data in each imputation are also helpful. 
Only the first imputation is shown here. The volume results have not 
hit their upper limits of 400, which is reassuring.
_imputation_=1
Variable Label N Mean Std Dev Minimum Maximum
dvol1
dvol2
dvol3
dvol4
dvol5
dvol6
dvar1
dvar2
dvar3
dvar4
dvar5
dvar6		 volume of offending sweep 1
volume of offending sweep 2
volume of offending sweep 3
volume of offending sweep 4
volume of offending sweep 5
volume of offending sweep 6
variety of offending sweep 1
variety of offending sweep 2
variety of offending sweep 3
variety of offending sweep 4
variety of offending sweep 5
variety of offending sweep 6		 4328
4328
4328
4328
4328
4328
4328
4328
4328
4328
4328
4328		 8.7381369
11.2189626
15.4582107
13.9671153
10.9198431
7.4226559
2.4976895
2.8669131
3.4690388
3.1446396
2.3572089
1.6039741		 12.9160252
20.0399975
22.7179659
20.5536305
16.1867972
12.0879224
2.6343264
3.0146144
3.3640472
3.2849235
2.7116128
2.3591945		 0
0
0
0
0
0
0
0
0
0
0
0		 129.0000000
312.0000000
386.0000000
288.0000000
225.0000000
121.0000000
16.0000000
17.0000000
19.0000000
19.0000000
18.0000000
19.0000000		 

We don't show details of all the checks that should be done to use the 
whole data set. 
No  further data manipulation was needed to get these variables to agree with each other,
since the bounds options in the fitting dealt with this.



back to top uparrow
  **** Now extract from PROC MIANALYZE that gives mean of each score. This for boys only. 






 Multiple Imputation Parameter Estimates - boys only




 Parameter
 Estimate
 Std Error
 95% Confidence Limits
 DF
 Minimum
 Maximum
 Theta0
 t for H0:
Parameter=Theta0		
 Pr > |t|






 dprev1
 0.792051
 0.008977
 0.77443
 0.80967
 921.54
 0.789402
 0.794427
 0
 88.23
 <.0001




 dprev2
 0.780082
 0.009068
 0.76230
 0.79787
 1842.6
 0.777067
 0.781179
 0
 86.02
 <.0001




 dprev3
 0.816172
 0.008442
 0.79962
 0.83273
 2777.9
 0.814070
 0.818182
 0
 96.67
 <.0001




 dprev4
 0.789767
 0.009087
 0.77192
 0.80761
 609.95
 0.788488
 0.793970
 0
 86.91
 <.0001




 dprev5
 0.743353
 0.013730
 0.71387
 0.77283
 13.839
 0.730471
 0.753769
 0
 54.14
 <.0001




 dprev6
 0.705619
 0.021761
 0.65289
 0.75835
 6.2537
 0.687072
 0.724075
 0
 32.43
 <.0001




 dvol1
 11.359029
 0.363039
 10.63921
 12.07885
 105.28
 11.201641
 11.590776
 0
 31.29
 <.0001




 dvol2
 13.376945
 0.472627
 12.45045
 14.30344
 7037.6
 13.315454
 13.470026
 0
 28.30
 <.0001




 dvol3
 17.158250
 0.521204
 16.13587
 18.18064
 1464.6
 17.000992
 17.270864
 0
 32.92
 <.0001




 dvol4
 16.105710
 0.506494
 15.11183
 17.09959
 1030.5
 16.037352
 16.310184
 0
 31.80
 <.0001




 dvol5
 13.231412
 0.702640
 11.63692
 14.82591
 8.818
 12.485495
 13.958612
 0
 18.83
 <.0001




 dvol6
 9.779373
 0.602592
 8.36739
 11.19136
 7.3304
 9.328909
 10.556637
 0
 16.23
 <.0001




 dvar1
 3.021380
 0.067738
 2.88776
 3.15500
 189.11
 2.995889
 3.056647
 0
 44.60
 <.0001




 dvar2
 3.320512
 0.070301
 3.18272
 3.45830
 28137
 3.308360
 3.325263
 0
 47.23
 <.0001




 dvar3
 3.837825
 0.079720
 3.68140
 3.99425
 1053.8
 3.813157
 3.857012
 0
 48.14
 <.0001




 dvar4
 3.600183
 0.079541
 3.44411
 3.75626
 1059.6
 3.583371
 3.628598
 0
 45.26
 <.0001




 dvar5
 2.769575
 0.100155
 2.55299
 2.98616
 12.875
 2.655094
 2.816811
 0
 27.65
 <.0001




 dvar6
 2.027044
 0.094958
 1.81505
 2.23904
 9.8571
 1.931932
 2.122430
 0
 21.35
 <.0001






*****  The rather small degrees of freedom for some estimates suggest that more imputations should have been run.
Multiple Imputation Parameter Estimates
Parameter Estimate Std Error 95% Confidence Limits DF Minimum Maximum Theta0 t for H0:
Parameter=Theta0		 Pr > |t|
GENDERFemale -0.182365 0.041632 -0.26892 -0.09581 21.084 -0.224780 -0.161694 0 -4.38 0.0003
SZINDEPManual_high_depr 0.023531 0.038324 -0.05349 0.10055 48.769 -0.008899 0.038867 0 0.61 0.5421
sectorBehavioural 0.333826 0.276691 -0.22171 0.88936 50.788 0.194677 0.522140 0 1.21 0.2332
sectorIndependent 0.174445 0.141173 -0.10283 0.45172 579.4 0.111849 0.197716 0 1.24 0.2171
sectorSpecial -0.968083 0.314235 -1.59229 -0.34388 90.807 -1.185340 -0.867281 0 -3.08 0.0027 


back to top uparrow
   
**** again detailed PROC MI output not shown. All variables had to sorted out - 
   see program file ex6sas.htm





back to top uparrow
   
   
**** Mean of each score for boys only. 
      From observed data only means for prevalence would be
    0.7895 0.7768 0.81100 0.7812 0.7274 0.6340, Now prevalence at sweep 
	6 a bit higher but all others very similar.
Same was true for the logistic regression.
Variable N Mean Std Dev Minimum Maximum
dvar1
dvol1
dvar2
dvol2
dvar3
dvol3
dvar4
dvol4
dvar5
dvol5
dvar6
dvol6
dprev1
dprev2
dprev3
dprev4
dprev5
dprev6		 21890
21890
21890
21890
21890
21890
21890
21890
21890
21890
21890
21890
21890
21890
21890
21890
21890
21890		 3.0243947
11.1868890
3.3406578
13.1315212
3.8855642
17.2696208
3.6707172
16.3657835
2.7711284
12.9211055
1.9160347
8.2956601
0.7943353
0.7833257
0.8160804
0.7919598
0.7492919
0.6772499		 2.8326314
14.8583493
3.2371275
21.7437621
3.6090678
23.7630851
3.5963630
23.1447190
2.9182549
17.8732482
2.2316574
12.7468642
0.4041957
0.4119882
0.3874275
0.4059150
0.4334306
0.4675387		 0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0		 16.0000000
129.0000000
17.0000000
312.0000000
18.0000000
245.0000000
18.0000000
288.0000000
16.0000000
149.0000000
16.0000000
121.0000000
1.0000000
1.0000000
1.0000000
1.0000000
1.0000000
1.0000000

Logistic regressions also give almost the same answers as observed data and the much smaller imputation.

Multiple Imputation Parameter Estimates
Parameter Estimate Std Error 95% Confidence Limits DF Minimum Maximum Theta0 t for H0:
Parameter=Theta0		 Pr > |t|
GENDERFemale -0.188723 0.033577 -0.25497 -0.12248 183.13 -0.198142 -0.173784 0 -5.62 <.0001
SZINDEPManual_high_depr 0.030178 0.037233 -0.04452 0.10488 52.42 0.011207 0.050687 0 0.81 0.4213
sectorBehavioural 0.252723 0.344938 -0.51136 1.01681 10.452 -0.032976 0.478006 0 0.73 0.4799
sectorIndependent 0.207273 0.207319 -0.25971 0.67426 9.2605 0.035661 0.377209 0 1.00 0.3428
sectorSpecial -0.976275 0.352051 -1.70360 -0.24895 23.557 -1.234814 -0.786841 0 -2.77 0.0107