topic Re: PROC MEANS in SAS Procedures

Re: PROC MEANS

data_null__ — Mon, 17 Mar 2014 16:05:31 GMT

Yes, but if you want examples you will need to be more specific.

Re: PROC MEANS

Idunnu — Mon, 17 Mar 2014 16:41:03 GMT

Suppose i want to subtract the mean of sbp in StdTyp 1 Period 1 Trtgrp 1 (highlighted) from the mean of StdTyp 1 Period 1 Trtgrp 2 (highlighted) to give a variable called X, what code do i use? I have pasted the output below. thanks. Tried different things but didnt work.

The MEANS Procedure

StdTyp Period Trtgrp N obs variable N mean Std Dev Minimum Maximum

sbp
dbp

118.7777778

73.9444444

16.4439203

8.9493684

95.0000000

58.0000000

156.0000000

90.0000000

sbp
dbp

110.8765432

72.3086420

23.4762554

15.2271484

138.0000000

90.0000000

sbp
dbp

117.2435897

77.2948718

13.7152734

7.8807821

92.0000000

60.0000000

148.0000000

94.0000000

sbp
dbp

118.9074074

73.4259259

18.7035145

9.8853844

94.0000000

56.0000000

168.0000000

96.0000000

sbp
dbp

107.6666667

68.1111111

11.9410316

5.6868154

92.0000000

59.0000000

130.0000000

77.0000000

sbp
dbp

121.4074074

75.5555556

14.7884689

9.8472612

86.0000000

60.0000000

148.0000000

102.0000000

sbp
dbp

121.7222222

77.3148148

12.7374306

8.0863829

98.0000000

64.0000000

149.0000000

94.0000000

sbp
dbp

104.1666667

67.8333333

4.6684170

4.5406659

98.0000000

62.0000000

114.0000000

78.0000000

Re: PROC MEANS

Idunnu — Wed, 19 Mar 2014 00:32:59 GMT

@data_null_;

Re: PROC MEANS

data_null__ — Wed, 19 Mar 2014 14:59:51 GMT

Based on the data set name and variables names I think perhaps you should ask about this as a statistical analysis question not just as a data summary. Perhaps could look at your data give some advice.

However I will show you how you can use PROC ANOVA to compute the means differences as per your example. Using PROC ANOVA allows you do get the differences directly without having to do much in the way of data manipulation.

data crossoverdesign;    
   infile datalines delimiter=',';  
   input Id StdTyp Period Trtgrp @;
   do d=1 to 3; 
      do r=1 to 3; 
         input sbp dbp @;
         if sbp eq 999 then sbp=.; 
         if dbp eq 999 then dbp=.; 
         output; 
         end; 
      end;   
   datalines;  
3,1,1,1,134,82,122,80,118,80,112,76,130,84,120,82,114,76,104,74,114,76  
3,1,2,2,113,61,116,72,115,67,122,70,136,76,128,72,122,72,126,72,123,71  
4,2,1,3,142,90,134,88,134,84,122,76,116,76,114,74,122,84,124,78,116,80  
4,2,2,1,128,86,133,81,135,81,129,83,116,74,130,80,138,84,140,80,134,74  
5,1,1,1,134,78,138,84,126,80,122,74,120,76,122,78,140,82,134,84,136,84  
5,1,2,2,126,76,122,74,116,76,134,82,126,88,130,84,132,78,128,82,126,78  
8,2,1,3,112,72,114,70,116,70,120,78,122,76,116,76,118,72,113,75,113,73  
8,2,2,1,116,72,106,76,110,72,125,75,131,73,126,78,110,68,109,67,105,71  
9,1,1,2,116,70,112,72,108,70,124,70,110,70,112,74,114,78,120,72,116,76  
9,1,2,1,102,60,104,64,104,66,110,74,107,77,112,74,104,64,113,75,102,70  
10,1,1,2,109,73,107,77,104,80,110,74,114,74,110,74,119,79,119,75,112,80    
10,1,2,1,118,78,117,83,118,82,108,72,106,70,98,70,117,81,110,78,110,80  
14,1,1,1,116,64,97,61,100,60,118,73,125,77,119,69,106,70,107,73,110,70  
14,1,2,2,114,70,110,72,112,76,100,66,113,73,100,68,116,74,112,74,112,72    
22,3,1,2,122,78,114,76,114,72,110,76,102,68,105,73,116,68,106,66,106,72    
22,3,2,3,102,64,112,74,120,80,112,74,100,78,116,72,95,67,100,70,100,70  
25,3,1,2,102,60,94,62,94,66,96,62,103,65,103,61,110,70,121,71,125,77    
25,3,2,3,98,60,101,65,96,64,99,59,101,65,98,60,102,62,96,68,104,62   
27,1,1,2,126,80,130,80,120,78,122,80,128,74,122,82,128,78,121,77,123,77    
27,1,2,1,116,76,120,72,120,74,134,84,136,88,128,84,136,86,136,82,132,80    
29,1,1,1,111,67,110,72,106,66,95,65,105,65,104,62,96,58,98,64,100,60    
29,1,2,2,109,71,112,72,110,72,94,62,96,62,98,60,108,70,102,70,100,66    
30,3,1,3,110,62,111,61,99,55,110,62,110,62,112,62,104,64,104,64,108,60  
30,3,2,2,116,68,114,64,114,64,102,60,100,56,106,56,116,66,108,62,108,64    
31,2,1,3,134,82,136,72,128,72,130,80,131,73,129,75,134,82,132,78,132,74    
31,2,2,1,122,74,128,78,124,80,126,78,126,72,128,78,126,82,132,80,136,80    
32,2,1,1,114,66,106,68,106,66,122,72,130,74,124,76,118,62,121,77,116,72    
32,2,2,3,104,62,103,67,102,66,109,69,106,66,104,66,114,70,108,70,112,72    
33,1,1,1,156,86,154,76,144,86,148,88,140,88,142,86,144,90,132,86,140,86    
33,1,2,2,168,94,146,96,154,92,157,89,158,90,158,90,133,83,142,84,146,88    
34,1,1,2,122,76,116,74,116,76,127,77,114,72,114,70,126,72,116,66,118,72    
34,1,2,1,126,80,126,80,120,82,113,71,116,74,108,74,110,72,109,71,110,78    
35,1,1,2,98,64,98,70,98,68,108,66,106,76,104,70,102,64,96,70,100,66  
35,1,2,1,999,999,999,999,999,999,106,74,98,76,96,76,98,66,97,71,92,62   
36,2,1,1,96,60,104,66,94,64,100,70,100,74,106,74,96,64,92,62,93,59   
36,2,2,3,100,72,104,74,108,78,103,63,104,62,98,64,98,64,98,64,100,72    
37,1,1,2,118,74,114,70,112,68,110,72,110,70,113,69,116,76,122,76,116,74    
37,1,2,1,126,84,127,83,120,82,126,78,125,77,118,80,106,64,104,68,110,68    
39,3,1,2,128,86,124,84,126,84,132,80,130,86,134,82,114,74,112,74,112,82    
39,3,2,3,123,79,120,82,124,84,106,68,102,72,102,72,108,76,112,74,106,76    
40,2,1,3,96,64,106,70,99,67,110,60,96,62,86,60,98,62,92,60,91,63  
40,2,2,1,121,77,118,76,116,82,104,68,107,69,108,76,99,75,98,64,98,70    
42,3,1,2,138,80,136,74,133,79,154,88,144,80,149,81,134,84,127,77,126,74    
42,3,2,3,126,76,125,75,121,73,130,76,118,80,118,68,132,74,118,70,126,74    
48,1,1,2,108,72,108,72,108,76,108,70,120,76,118,78,0,0,0,0,0,0    
48,1,2,1,106,74,108,70,108,70,111,71,112,78,116,78,104,70,100,68,101,65    
49,3,1,3,90,62,94,64,86,60,90,60,92,60,86,56,86,58,87,59,90,60    
49,3,2,2,88,64,88,60,92,66,86,60,94,60,92,64,88,56,92,62,92,62    
50,1,1,1,114,68,106,66,106,70,100,66,100,70,105,63,112,64,98,64,110,64  
50,1,2,2,98,62,96,58,96,56,106,66,98,66,99,63,105,67,102,58,100,62   
52,2,1,3,129,67,120,68,130,72,121,67,118,66,114,68,114,70,111,67,114,68    
52,2,2,1,118,70,120,70,106,72,112,68,116,68,118,70,108,66,108,68,108,68    
53,1,1,2,112,84,112,82,114,84,110,76,110,74,112,78,112,88,114,84,114,84    
53,1,2,1,120,82,124,86,112,82,140,88,131,91,128,92,112,78,114,88,121,87    
57,1,1,2,124,82,132,84,132,90,118,70,120,80,110,72,130,80,138,80,131,79    
57,1,2,1,136,88,136,86,142,90,148,88,142,90,140,94,142,72,138,84,148,84    
58,3,1,2,94,74,98,66,86,68,102,70,98,66,90,64,86,64,96,66,94,66   
58,3,2,3,102,66,100,74,96,68,96,70,103,73,99,73,102,74,100,68,94,68  
61,2,1,3,144,92,139,83,142,90,134,92,138,88,134,90,136,90,148,102,142,92   
61,2,2,1,146,94,144,92,149,93,138,92,126,90,128,92,134,90,132,90,124,88    
62,3,1,3,104,64,100,66,96,64,100,66,100,70,103,67,104,70,102,66,102,68  
62,3,2,2,106,70,108,68,98,74,106,70,102,68,102,70,103,65,100,64,96,64   
;;;;
   run;  
proc sort data=crossoverdesign;
   by stdtyp period trtgrp;
   run; 
proc anova data=crossoverdesign;
   by stdtyp period;
   class trtgrp;
   model sbp dbp = trtgrp;
   means trtgrp / cldiff dunnett; 
   ods select none; 
   ods output CLDiffs=CLDiffs(drop=method uppercl lowercl significance);
   run; 
ods select all; 
proc print data=CLDiffs;
   run; 

Re: PROC MEANS

SteveDenham — Wed, 19 Mar 2014 17:49:29 GMT

Follow the advice of data_null_; This will give you the differences in the raw means you are looking for.

Then start thinking about things like cross-over designs, repeated measures on subjects, unbalancedness, and the fact that sbp (systolic blood pressure?) and dbp (diastolic blood pressure) are surely correlated within subject as well. A full description of the experimental design would lead to a more comprehensive analysis.

My biggest problem with unbalanced designs is that the difference between raw means is a biased estimator of effects.

Steve Denham

Re: PROC MEANS

Idunnu — Wed, 19 Mar 2014 18:04:10 GMT

TThanks! @data_null_; .... I appreciate your responses!!!

Re: PROC MEANS

Idunnu — Wed, 19 Mar 2014 19:35:35 GMT

yea, its actually a crossover design analysis

Re: PROC MEANS

SteveDenham — Thu, 20 Mar 2014 12:16:45 GMT

Then the raw means are relatively meaningless. Google is your friend--there are multiple examples out there of the analysis of cross-over designs using PROC GLM, PROC MIXED and PROC GLIMMIX. The lsmeans generated from these analyses accommodate the multiple effects, and the imbalance seen. I would really suggest going that way with this data.

Steve Denham

Re: PROC MEANS

Idunnu — Thu, 20 Mar 2014 14:24:43 GMT

YEa, I ended up using the proc mixed method but I'll need the raw means just to get a description of the two variables.. Thank you!!