I came across this macro for calculating various effect size statistics, but I am unsure how to use it. Can someone show me how I would input in my data and run this?
For example, I would like to compute Cohen's d and Hedges' g. My values are the following:
NE = 103
NC = 100
ME = 1.44
MC = 1.43
SE = 0.65
SC = 0.59
%macro compeff(indata,outdata);
/******************************************************/
/* INDATA: Name of the Input Data Set */
/* OUTDATA: Name of the Output Data Set */
/******************************************************/
data &outdata; set &indata;
/******************************************************/
/* INPUT DATA: */
/* NE: Sample size for Experimental Group */
/* NC: Sample size for Control Group */
/* ME: Treatment Effect for Experimental Group */
/* MC: Treatment Effect for Control Group */
/* SE: Standard Deviation for Experimental Group */
/* SC: Standard Deviation for Control Group */
/* */
/* OUTPUT DATA: */
/* TYPE: 0 = Glass' Delta */
/* 1 = Cohen's d */
/* 2 = Hedges' g */
/* 3 = Hedges' gu */
/* 4 = Transformed Correlation rpb */
/* ESTIMATE: Effect-Szie Estimate */
/* STD: Estimated Standard deviation */
/* LOWER: Lower Bound of 95% Confidence Interval */
/* UPPER: Upper Bound of 95% Confidence Interval */
/******************************************************/
dfd = ne+nc;
dfg = dfd-2;
/******************************************************/
/* Compute Glass's Delta */
/******************************************************/
del = (me-mc)/sc;
vdel = dfd/(ne*nc)+(del*del)/(2*nc-2);
sdel = sqrt(vdel);
/******************************************************/
/* Compute Cohen's d */
/******************************************************/
sm = sqrt(((ne-1)*(se*se)+(nc-1)*(sc*sc))/dfd);
dd = (me-mc)/sm;
vdd = dfd/(ne*nc)+0.5*dd*dd/dfg;
sdd = sqrt(vdd);
/******************************************************/
/* Compute Hedges' g */
/******************************************************/
sp = sqrt(((ne-1)*(se*se)+(nc-1)*(sc*sc))/dfg);
gg = (me-mc)/sp;
vgg = dfd/(ne*nc)+0.5*gg*gg/dfg;
sgg = sqrt(vgg);
/******************************************************/
/* Compute Hedges' gu */
/******************************************************/
corrf=exp(lgamma(0.5*dfg)-lgamma(0.5*(dfg-1))
-log(sqrt(0.5*dfg)));
gu = gg * corrf;
vgu = dfd/(ne*nc)+0.5*gu*gu/dfd;
sgu = sqrt(vgu);
/*********************************************************/
/* Compute Point-Biserial Correlation */
/*********************************************************/
rpb = sqrt((ne*nc*dd*dd)/(ne*nc*dd*dd+dfd*dfd));
zrpb = 0.5*log((1+rpb)/(1-rpb));
vzrpb = 1 / (dfd-3);
szrpb = sqrt(vzrpb);
type = 0;
estimate = del;
std = sdel;
lower = del + probit(0.025) * sdel;
upper = del + probit(0.975) * sdel;
output;
type = 1;
estimate = dd;
std = sdd;
lower = dd + probit(0.025) * sdd;
upper = dd + probit(0.975) * sdd;
output;
type = 2;
estimate = gg;
std = sgg;
lower = gg + probit(0.025) * sgg;
upper = gg + probit(0.975) * sgg;
output;
type = 3;
estimate = gu;
std = sgu;
lower = gu + probit(0.025) * sgu;
upper = gu + probit(0.975) * sgu;
output;
type = 4;
estimate = rpb;
std = szrpb;
lower = zrpb + probit(0.025) * szrpb;
upper = zrpb + probit(0.975) * szrpb;
lower = (exp(2*lower)-1)/(exp(2*lower)+1);
upper = (exp(2*upper)-1)/(exp(2*upper)+1);
output;
keep ne nc me mc se sc type estimate lower upper;
%mend compeff;
You need to create a dataset according to the description in the comment section:
/* INPUT DATA: */
/* NE: Sample size for Experimental Group */
/* NC: Sample size for Control Group */
/* ME: Treatment Effect for Experimental Group */
/* MC: Treatment Effect for Control Group */
/* SE: Standard Deviation for Experimental Group */
/* SC: Standard Deviation for Control Group */
Like
data in;
input ne nc me mc se sc;
datalines;
103 100 1.44 1.43 0.65 0.59
;
You need to create a dataset according to the description in the comment section:
/* INPUT DATA: */
/* NE: Sample size for Experimental Group */
/* NC: Sample size for Control Group */
/* ME: Treatment Effect for Experimental Group */
/* MC: Treatment Effect for Control Group */
/* SE: Standard Deviation for Experimental Group */
/* SC: Standard Deviation for Control Group */
Like
data in;
input ne nc me mc se sc;
datalines;
103 100 1.44 1.43 0.65 0.59
;
Are you ready for the spotlight? We're accepting content ideas for SAS Innovate 2025 to be held May 6-9 in Orlando, FL. The call is open until September 25. Read more here about why you should contribute and what is in it for you!
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.