topic Re: Calculate adjusted mean by using proc mixed in Statistical Procedures

Calculate adjusted mean by using proc mixed

Aryyyan — Wed, 26 May 2021 23:34:47 GMT

Hello everyone,

I want to calculate adjusted mean by using proc mixe as in Table 12 (please see attached mock shells). I have attached my sas file and data file.

Thank you

Re: Calculate adjusted mean by using proc mixed

PaigeMiller — Thu, 27 May 2021 13:39:26 GMT

Many of us will not download files (especially Microsoft Office files) as they can be a security threat.

Please provide your code by copying it as text and pasting it into the box that appears when you click on the "little running man" icon. Please provide a portion of your data following these instructions: https://blogs.sas.com/content/sastraining/2016/03/11/jedi-sas-tricks-data-to-data-step-macro/

Please explain what "adjusted mean" you are looking for. And please explain what problems you are having with your code.

Re: Calculate adjusted mean by using proc mixed

ballardw — Thu, 27 May 2021 15:59:10 GMT

When you provide spreadsheets as "data" we have to make decisions to turn it into a SAS data set.

My decisions are not likely to be yours. So I may not have variables of the same name or type as your SAS data set.

So besides the security risks or organization policies of not opening files from unknown sources I don't open spreadsheets as "data".

Instructions here: https://communities.sas.com/t5/SAS-Communities-Library/How-to-create-a-data-step-version-of-your-data-AKA-generate/ta-p/258712 will show how to turn an existing SAS data set into data step code that can be pasted into a forum code box using the </> or "running man" icon or attached as text to show exactly what you have and that we can test code against.

Re: Calculate adjusted mean by using proc mixed

jiltao — Thu, 27 May 2021 17:21:48 GMT

What you wanted might be obtained by using the LSMEANS statement in PROC MIXED --

lsmeans avisitn*trt01p;

The result from this LSMEANS statement would depend on your PROC MIXED model. The independent variables in your model statement would include avisitn trt01p avisitn*trt01p and other effects that you will have to decide what to include. You also need to specify a CLASS statement for your classification variables, and a RANDOM or a REPEATED statement to model the correlations in the repeated measures. You might find an example in the PROC MIXED documentation helpful --

https://go.documentation.sas.com/doc/en/pgmsascdc/v_012/statug/statug_mixed_examples02.htm

Jill

Re: Calculate adjusted mean by using proc mixed

Aryyyan — Thu, 27 May 2021 21:51:37 GMT

Hi everyone,

Thank you for your reply. My data is as below.

USUBJID	AGegr1	Trt01p	AVALC	NEREASN	trtN
LRN-12-6007	<65	Placebo	NE	Death Before Measurement	2
LRN-12-6007	<65	Total	NE	Death Before Measurement	3
LRN-12-6009	<65	CMP123	PR		1
LRN-12-6009	<65	Total	PR		3
LRN-12-6151	<65	Placebo	NE	Death Before Measurement	2
LRN-12-6151	<65	Total	NE	Death Before Measurement	3
LRN-12-6153	<65	CMP123	NE	Droped Study	1
LRN-12-6153	<65	Total	NE	Droped Study	3
LRN-12-6155	>=65	Placebo	NE	Withdrown Consent	2
LRN-12-6155	>=65	Total	NE	Withdrown Consent	3
LRN-12-6157	<65	CMP123	NE	Withdrown Consent	1
LRN-12-6157	<65	Total	NE	Withdrown Consent	3
LRN-12-6159	<65	Placebo	PR		2
LRN-12-6159	<65	Total	PR		3
LRN-12-6161	<65	CMP123	PR		1
LRN-12-6161	<65	Total	PR		3
LRN-12-6163	<65	Placebo	NE	Death Before Measurement	2
LRN-12-6163	<65	Total	NE	Death Before Measurement	3
LRN-12-6165	>=65	CMP123	CR		1
LRN-12-6165	>=65	Total	CR		3

I calculated unadjusted mean as below.

proc sql noprint;
select count (distinct usubjid) into : trtX from adrs2 where trtN = 1;
select count (distinct usubjid) into : trtY from adrs2 where trtN = 2;
select count (distinct usubjid) into : trtZ from adrs2 where trtN = 3;
%put &trtX &trtY &trtZ;
run;

proc freq data = adrs2;
tables trtN*agegr1/out=dummy (drop=percent);
tables avalc*trtN*agegr1 / out = adrs3 (drop=percent);
tables avalc*trtN*nereasn*agegr1 / out=utdtype (drop=percent);
run;

data adrs4 (drop=avalc nereasn) ;
length avalc $40.;
set adrs3 (in=a) utdtype (in=b);
by avalc;
if nereasn ne " " then _avalc = nereasn;
else if nereasn eq " " then _avalc = avalc;
rename _avalc = avalc;
run;

proc sort data=adrs4 out = adrs5 nodupkey;
by trtn count avalc;
run;

data adrs6;
set adrs5;
if trtN = 1 then denom = &trtX;
if trtN = 2 then denom = &trtY;
if trtN = 3 then denom = &trtZ;
percent = put ((count/denom)*100, 5.1);
CP = count||" ("||percent||")";
drop count denom percent;
run;

*_______________________________________________________________________________________________;

data ORR;
set adrs2;
if avalc in ("CR", "PR") then rate = 1;
else rate = 2;
run;

proc sort data=orr;
by agegr1 trtn;
run;

ods output OneWayFreqs = ORRfreq(rename = (frequency = count))
BinomialCLs = limit (where=(Type = "Clopper-Pearson (Exact)"))
BinomialTest = test;
proc freq data=orr order=formatted;
by agegr1 trtn;
tables rate/binomial(ac wilson exact) nocol norow nopercent;
run;
ods output close;

data limit2;
set limit;
CP = "("||strip(put(LowerCL,5.2))||", "||strip(put(UpperCL,5.2)) ||")";
keep agegr1 CP trtN;
run;

data orrfreq1;
set orrfreq;
if rate = 1 then do;
if trtN = 1 then denom = &trtX;
if trtN = 2 then denom = &trtY;
if trtN = 3 then denom = &trtZ;
CD = count||"/"||left(denom);
percent = put ((count/denom)*100, 5.1);
CP = strip (CD)||" ("||strip(percent)||"%)";
end;
if rate = 2 then delete;
keep agegr1 cp trtN;
run;

*_____________________________________________________________________________________________;
proc sort data=dummy out=dummy1 (keep=trtN agegr1) nodupkey; *to insert blank line for table;
by agegr1 trtN;
run;

proc sort data=adrs6 out=adrs7;
by agegr1 trtN;
run;

proc sort data=limit2 out=limit3;
by agegr1 trtN;
run;

proc sort data=orrfreq1 out=orrfreq2;
by agegr1 trtN;
run;

data adrs8;
length CP $50 avalc $50;
set adrs7 (in=a) dummy1 (in=b) orrfreq2 (in=c) limit3 (in=d);
by agegr1 trtN;
if b then do;
avalc = "BEST OVERALL RESPONSE";
end;
if c then do;
avalc = "OBJECTIVE RESPONSE RATE (1)";
end;
if d then do;
avalc = " (95% CI)";
end;
run;

data adrs9;
length avalc1 $ 50;
set adrs8;
if avalc = "BEST OVERALL RESPONSE" then do;
AVALC1 = "BEST OVERALL RESPONSE";
avaln = 1;
end;
if AVALC = "CR" then do
AVALC1 = "COMPLETE RESPONSE (CR)";
avaln =2;
end;
if AVALC = "PR" then do;
AVALC1 = "PARTIAL RESPONSE (PR)";
avaln =3;
end;
if AVALC = "SD" then do;
AVALC1 = "STABLE DISEASE (SD)";
avaln =4;
end;
if AVALC = "PD" then do;
AVALC1 = "PROGRESSIVE DISEASE (PD)";
avaln =5;
end;
if AVALC = "NE" then do;
AVALC1 = "UNABLE TO DETERMINE (UTD)";
avaln =6;
end;
if AVALC = "Death Before Measurement" then do;
AVALC1 = " REASON UTD 1";
avaln =7;
end;
if AVALC = "Droped Study" then do;
AVALC1 = " REASON UTD 2";
avaln =8;
end;
if AVALC = "Withdrown Consent" then do;
AVALC1 = " REASON UTD X";
avaln =9;
end;
if AVALC = "OBJECTIVE RESPONSE RATE (1)" then do;
AVALC1 = "OBJECTIVE RESPONSE RATE (1)";
avaln =91;
end;
if AVALC = " (95% CI)" then do;
AVALC1 = " (95% CI)";
avaln =92;
end;
drop avalc;
rename avalc1 = avalc;

RUN;

proc sort data = adrs9 out= adrs10;
by agegr1 avaln avalc;
run;

proc transpose data = adrs10 out = adrs11 (drop = _NAME_) prefix=TRT;
by agegr1 avaln avalc;
id trtN;
var cp;
run;

data adrs12;
set adrs11;
by agegr1;
if agegr1 = "<65" then agegr1n = 1;
else agegr1n = 2;
run;

data adrs13;
set adrs12;
if trt1 = " " and avalc ne "BEST OVERALL RESPONSE" then trt1 = " 00 (00.0)";
if trt2 = " " and avalc ne "BEST OVERALL RESPONSE" then trt2 = " 00 (00.0)";
if trt3 = " " and avalc ne "BEST OVERALL RESPONSE" then trt3 = " 00 (00.0)";
run;

Now, I have to calculate summary of adjusted mean of change from baseline viral load at week 12 and 16 by using proc mixed stage as factor and treatment.

Re: Calculate adjusted mean by using proc mixed

PaigeMiller — Fri, 28 May 2021 10:52:33 GMT

Hello, @Aryyyan , specific instructions were given on how to present a portion of your data: https://blogs.sas.com/content/sastraining/2016/03/11/jedi-sas-tricks-data-to-data-step-macro/

Other questions have not been answered, such as what is wrong with your code now, and what "adjusted mean" do you want.

Re: Calculate adjusted mean by using proc mixed

SteveDenham — Fri, 28 May 2021 12:04:42 GMT

Echoing @PaigeMiller here, could you provide your definition of "adjusted" mean. I can think of several ways to adjust - subtract baseline, use baseline as a covariate, calculate marginal means to accommodate unequal sample sizes, weight factors in the model by (number of observations, 1/standard deviation, 1/variance, custom weights, any of these in conjunction with the first two). Since I am not really familiar with standards in this field, I don't know how to advise.you.

SteveDenham