BookmarkSubscribeRSS Feed
ambadi007
Quartz | Level 8

HI Team,

I have a requirement to create Pvalue based on distribution of data , below is the requirement.

1 If data follows normal distribution, P-values were derived using paired sample t-test else by Wilcoxon test for baseline vs all the timepoints.

The below codes i have written , but the codes are too lengthy. Could you please help to make it in a macro

/*Method 1 - Paired T test*/

%macro pir(dat1=,visn=,parn=,out=,rm=);

data &dat1;
set input;
where parcat1n=&parn. and  avisn=&visn.;
run;

ods trace on;
ods output TTests=&out. ;
proc ttest data=&dat1. alpha=0.05;
paired base*aval;
run; 
ods trace off;

data &out._1(rename=(probc=&rm.));
set &out.;
probc=put(Probt,pvalue6.4)||"^{Super 1}";
label=" P-value";
ord=8.1;
keep label ord Variable1 Variable2 Difference probc;
run;

%mend;
%pir(dat1=input1,visn=1,parn=1,out=l_wk1,rm=_Lesional_);
%pir(dat1=input1,visn=1,parn=2,out=nl_wk1,rm=_Non_Lesional_);
%pir(dat1=input1,visn=1,parn=3,out=h_wk1,rm=_Healthy_);

%pir(dat1=input1,visn=2,parn=1,out=l_wk2,rm=_Lesional_);
%pir(dat1=input1,visn=2,parn=2,out=nl_wk2,rm=_Non_Lesional_);
%pir(dat1=input1,visn=2,parn=3,out=h_wk2,rm=_Healthy_);

%pir(dat1=input1,visn=3,parn=1,out=l_wk3,rm=_Lesional_);
%pir(dat1=input1,visn=3,parn=2,out=nl_wk3,rm=_Non_Lesional_);
%pir(dat1=input1,visn=3,parn=3,out=h_wk3,rm=_Healthy_);

%pir(dat1=input1,visn=4,parn=1,out=l_wk4,rm=_Lesional_);
%pir(dat1=input1,visn=4,parn=2,out=nl_wk4,rm=_Non_Lesional_);
%pir(dat1=input1,visn=4,parn=3,out=h_wk4,rm=_Healthy_);

%pir(dat1=input1,visn=5,parn=1,out=l_wk6,rm=_Lesional_);
%pir(dat1=input1,visn=5,parn=2,out=nl_wk6,rm=_Non_Lesional_);
%pir(dat1=input1,visn=5,parn=3,out=h_wk6,rm=_Healthy_);

%pir(dat1=input1,visn=6,parn=1,out=l_wk8,rm=_Lesional_);
%pir(dat1=input1,visn=6,parn=2,out=nl_wk8,rm=_Non_Lesional_);
%pir(dat1=input1,visn=6,parn=3,out=h_wk8,rm=_Healthy_);

%pir(dat1=input1,visn=7,parn=1,out=l_wk12,rm=_Lesional_);
%pir(dat1=input1,visn=7,parn=2,out=nl_wk12,rm=_Non_Lesional_);
%pir(dat1=input1,visn=7,parn=3,out=h_wk12,rm=_Healthy_);

%pir(dat1=input1,visn=8,parn=1,out=l_wk16,rm=_Lesional_);
%pir(dat1=input1,visn=8,parn=2,out=nl_wk16,rm=_Non_Lesional_);
%pir(dat1=input1,visn=8,parn=3,out=h_wk16,rm=_Healthy_);

%pir(dat1=input1,visn=9,parn=1,out=l_wk28,rm=_Lesional_);
%pir(dat1=input1,visn=9,parn=2,out=nl_wk28,rm=_Non_Lesional_);
%pir(dat1=input1,visn=9,parn=3,out=h_wk28,rm=_Healthy_);


%macro mrg(in_1=,in_2=,in_3=,vsn=,out=);
data &out.;
merge &in_1. &in_2. &in_3.;
by Variable1 Variable2;
avisn=&vsn.;
run;
%mend;

 %mrg(in_1=l_wk1_1,in_2=nl_wk1_1,in_3=h_wk1_1,vsn=1,out=p_wk1);
 %mrg(in_1=l_wk2_1,in_2=nl_wk2_1,in_3=h_wk2_1,vsn=2,out=p_wk2);
 %mrg(in_1=l_wk3_1,in_2=nl_wk3_1,in_3=h_wk3_1,vsn=3,out=p_wk3);
 %mrg(in_1=l_wk4_1,in_2=nl_wk4_1,in_3=h_wk4_1,vsn=4,out=p_wk4);
 %mrg(in_1=l_wk6_1,in_2=nl_wk6_1,in_3=h_wk6_1,vsn=5,out=p_wk6);
 %mrg(in_1=l_wk8_1,in_2=nl_wk8_1,in_3=h_wk8_1,vsn=6,out=p_wk8);
 %mrg(in_1=l_wk12_1,in_2=nl_wk12_1,in_3=h_wk12_1,vsn=7,out=p_wk12);
 %mrg(in_1=l_wk16_1,in_2=nl_wk16_1,in_3=h_wk16_1,vsn=8,out=p_wk16);
 %mrg(in_1=l_wk28_1,in_2=nl_wk28_1,in_3=h_wk28_1,vsn=9,out=p_wk28);

 /*Paired Ttest values*/

data PT_pval(rename=(variable2=varname));
set p_wk1 p_wk2 p_wk3 p_wk4 p_wk6 p_wk8 p_wk12 p_wk16 p_wk28;
run;

/*Checking normality of the dristribution to understand which test needs to perform*/
 
%macro uni(var=,a=,b=,d=,rnm=);

data input_1;
set input;
where  parcat1n=&b. ;
run;

proc sort;by avisn;run;

ods trace on;
ods output TestsForNormality=&d.;
Proc univariate data=input_1 normal;
var &var.;
by avisn;
RUN;
ods trace off;

data &d.(rename=(pValue=&rnm.));
set &d.;
where TEST="Shapiro-Wilk";
run;

%mend;

%uni(var=chg,b=1,d=pv_l_0s,rnm=ls);

%uni(var=chg,b=2,d=pv_nl_0s,rnm=nls);

%uni(var=chg,b=3,d=pv_h_0s,rnm=hs);


%macro mr(a=,in=,ord=);
data &a.;
merge &in.;
by avisn;
sord=&ord.;
keep avisn  VarName Test ls nls hs sord;
run;
%mend;
%mr(a=pval_0s,in=pv_l_0s pv_nl_0s pv_h_0s,ord=0);


data pvalue_T415;
	Length nvar $9.;
	set pval_0s;
	if ls<0.05 then nvar="ls";
	if nls<0.05 then nvar="nls";
	if hs<0.05 then nvar="hs";
	if ls<0.05 and hs<0.05 then nvar="lshs";
	if ls<0.05 and nls<0.05 then nvar="lsnls";
	if nls<0.05 and hs<0.05 then nvar="nlshs";
	if ls<0.05 and nls<0.05 and hs<0.05 then nvar="lsnlshs";
	if ls ne . and nls ne . and hs ne .;
run;

/*Method 2 Non parametric method*/

%macro par(a=,b=,c=,rnm=);
data input1;
set input;
where parcat1n=&a.  and avisn in &b.;
run;;

proc sort data=input1;
by parcat1;
run;

ods trace on;
ods output TestsForLocation=&c.;
Proc univariate data=input1;
Var chg;
Run;
ods trace off;

data &c.(rename=(pValue=&rnm.));
Length varname $4.;
set &c.;
where Test="Signed Rank";
varname="AVAL";
run;

%mend;
%par(a=1,b=(0,1),c=ls_1,rnm=npls);
%par(a=2,b=(0,1),c=nls_1,rnm=npnls);
%par(a=3,b=(0,1),c=hs_1,rnm=nphs);

%par(a=1,b=(0,2),c=ls_2,rnm=npls);
%par(a=2,b=(0,2),c=nls_2,rnm=npnls);
%par(a=3,b=(0,2),c=hs_2,rnm=nphs);

%par(a=1,b=(0,3),c=ls_3,rnm=npls);
%par(a=2,b=(0,3),c=nls_3,rnm=npnls);
%par(a=3,b=(0,3),c=hs_3,rnm=nphs);

%par(a=1,b=(0,4),c=ls_4,rnm=npls);
%par(a=2,b=(0,4),c=nls_4,rnm=npnls);
%par(a=3,b=(0,4),c=hs_4,rnm=nphs);

%par(a=1,b=(0,5),c=ls_5,rnm=npls);
%par(a=2,b=(0,5),c=nls_5,rnm=npnls);
%par(a=3,b=(0,5),c=hs_5,rnm=nphs);

%par(a=1,b=(0,6),c=ls_6,rnm=npls);
%par(a=2,b=(0,6),c=nls_6,rnm=npnls);
%par(a=3,b=(0,6),c=hs_6,rnm=nphs);

%par(a=1,b=(0,7),c=ls_7,rnm=npls);
%par(a=2,b=(0,7),c=nls_7,rnm=npnls);
%par(a=3,b=(0,7),c=hs_7,rnm=nphs);


%par(a=1,b=(0,8),c=ls_8,rnm=npls);
%par(a=2,b=(0,8),c=nls_8,rnm=npnls);
%par(a=3,b=(0,8),c=hs_8,rnm=nphs);

%par(a=1,b=(0,9),c=ls_9,rnm=npls);
%par(a=2,b=(0,9),c=nls_9,rnm=npnls);
%par(a=3,b=(0,9),c=hs_9,rnm=nphs);

%macro mr(a=,in=,ord=);
data &a.;
merge &in.;
by varname;
sord=&ord.;
keep varname test npls npnls nphs sord;
run;
%mend;

%mr(a=np_1,in=ls_1 nls_1 hs_1,ord=1);

%mr(a=np_2,in=ls_2 nls_2 hs_2,ord=2);

%mr(a=np_3,in=ls_3 nls_3 hs_3,ord=3);

%mr(a=np_4,in=ls_4 nls_4 hs_4,ord=4);

%mr(a=np_5,in=ls_5 nls_5 hs_5,ord=5);

%mr(a=np_6,in=ls_6 nls_6 hs_6,ord=6);

%mr(a=np_7,in=ls_7 nls_7 hs_7,ord=7);

%mr(a=np_8,in=ls_8 nls_8 hs_8,ord=8);

%mr(a=np_9,in=ls_9 nls_9 hs_9,ord=9);


data np_T415(rename=(ls=npls nls=npnls hs=nphs sord=avisn));
Length varname $4.;
set np_1-np_9;
ls=put(npls,pvalue6.4)||"^{Super 2}";
nls=put(npnls,pvalue6.4)||"^{Super 2}";
hs=put(nphs,pvalue6.4)||"^{Super 2}";
drop npls npnls nphs;
run;

proc sort data=np_T415;by avisn;run;

proc sort data=pvalue_T415;by avisn;run;

data final_T415;
	Length varname $4.;
	merge pvalue_T415(in=a) np_T415(in=b);
	by avisn;
	if a;
run;

/*Merging both Paired T test vlaue and non parameter value*/

data fi_pval;
	merge final_T415(in=a) PT_pval(in=b);
	by avisn;
	if a;
run;

/*Assinging Pavlues based on the requiremnts- Less than 0.05 for Wilcoxon and rest are assigned with T test*/

data final_pval;
set fi_pval;
	if nvar="ls" then _Lesional_=npls;
	if nvar="nls" then _Non_Lesional_=npnls;
	if nvar="hs" then _Healthy_=nphs;
	if nvar="lshs" then _Healthy_=nphs;
	if nvar="lshs" then _Lesional_=npls;
	if nvar="nlshs" then _Healthy_Skin_Area=nphs;
	if nvar="nlshs" then _Non_Lesional_=npnls;
	if nvar="lsnls" then _Non_Lesional_=npnls;
	if nvar="lsnls" then _Lesional_=npls;
	if nvar="lsnlshs" then  _Lesional_=npls;
	if nvar="lsnlshs" then _Non_Lesional_=npnls;
	if nvar="lsnlshs" then _Healthy_=nphs;
	;
run;

, Is there any way to make it small and create a macro for this

5 REPLIES 5
PaigeMiller
Diamond | Level 26

I have a requirement to create Pvalue based on distribution of data , below is the requirement. 1 If data follows normal distribution, P-values were derived using paired sample t-test else by Wilcoxon test for baseline vs all the timepoints. The below codes i have written , but the codes are too lengthy. Could you please help to make it in a macro

 

Not able to test this, as I don't have your data. But the general idea is to sort and then use a BY statement.

 

proc sort data=input;
     by parcat1n avisn;
run;

proc ttest data=input alpha=0.05;
    ods output ttests=out;
    by parcat1n avisn;
    paired base*aval;
run;

 

The general idea here (and hopefully for the rest of your SAS life) is to avoid using macros when a BY statement(s) can do the work.

 

Also, the same idea applies to using testing the distribution and for non-parametric tests, no need for a macro there either, so I did not write code for that, see if you can follow what I did to get distribution tests and non-parametric tests working without a macro.

 

 

--
Paige Miller
ambadi007
Quartz | Level 8
As you suggested the first method I have changed and it works...but in second method i need to show baseline VS all other timepoints (i.e visits)
PaigeMiller
Diamond | Level 26

@ambadi007 wrote:
...but in second method i need to show baseline VS all other timepoints (i.e visits)

I'm sorry but I don't know what this brief comment means, I don't know what doesn't work, I don't even know what baseline you are referring to. Please explain in detail.

--
Paige Miller
ambadi007
Quartz | Level 8
need to check the Pvalue for first visit(Baseline) and all other visits
PaigeMiller
Diamond | Level 26

@ambadi007 wrote:
need to check the Pvalue for first visit(Baseline) and all other visits

Hi, I find this to be unclear as well, and request a detailed explanation. Be generous with information!

--
Paige Miller

SAS Innovate 2025: Save the Date

 SAS Innovate 2025 is scheduled for May 6-9 in Orlando, FL. Sign up to be first to learn about the agenda and registration!

Save the date!

How to Concatenate Values

Learn how use the CAT functions in SAS to join values from multiple variables into a single value.

Find more tutorials on the SAS Users YouTube channel.

SAS Training: Just a Click Away

 Ready to level-up your skills? Choose your own adventure.

Browse our catalog!

Discussion stats
  • 5 replies
  • 797 views
  • 0 likes
  • 2 in conversation