topic Levene's test for equality of variances for multiple independent variable in Statistical Procedures

Levene's test for equality of variances for multiple independent variable

abdulla — Sat, 16 Feb 2019 01:49:36 GMT

Hi,

I know that the following is applicable for equal variance levene's test.

Proc glm data=want;

class age;

model y=age;

means age / hovtest;

run;

But if I have two variables with interaction term, what will be the code for levene's test? I tried the following but it didn't work.

Proc glm data=want;

class age sex;

model y=age sex age*sex;

means age sex age*sex/ hovtest;

run;

Please help

Re: Levene's test for equality of variances for multiple independent variable

PGStats — Sat, 16 Feb 2019 06:09:31 GMT

The documentation states "Note that this option is ignored unless your MODEL statement specifies a simple one-way model."

You may test the homogeneity of variance across all combined classes by creating a new variable:

data heartTest;
set sashelp.heart;
if cmiss(smoking, sex) = 0 then smokeSex = cats(smoking, sex);
run;

proc glm data=heartTest plots=none;
class smokeSex;
model cholesterol = smokeSex;
means smokeSex / hovtest=bf;
run;

Re: Levene's test for equality of variances for multiple independent variable

abdulla — Sat, 16 Feb 2019 06:30:25 GMT

Hey PG you are awesome. Thanks a lot

Re: Levene's test for equality of variances for multiple independent variable

Kurt_Bremser — Sat, 16 Feb 2019 09:11:03 GMT

I moved the question to the Statistical Procedures community.

"Community Matters" is for topics concerning the working of the Communities website.