How to do a Test of Homogeneity of Variance in SAS EG

[davidvalentine]

Hello Beautiful People!

I hope all is well!  I am trying to see if the variances of my samples are the same (for brevity's sake I only included 6 data, but my data sets in question have hundreds of rows), but I am not sure which procedure to use in SAS EG 8.1.  Also, is there a way to do this using the "Tasks" feature in SAS EG?  The sample data are below:

data work.DummySales;
infile datalines;
	input 
		Month monyy6.
		Sales: dollar10.2 SalesB: dollar10.2 SalesC: dollar10.2;
		format Month monyy. Sales dollar10.2 SalesB dollar10.2 SalesC dollar10.2;
;
datalines;
Jan-12  $6,300.28  $5,400.12  $6,002.77
Feb-12  $7,298.56  $7,919.22  $7,491.08
Mar-12  $5,098.79  $5,989.01  $5,488.23
Apr-12  $5,971.96  $5,622.71  $6,122.21
May-12  $5,585.79  $5,588.66  $5,771.39
Jun-12  $5,575.80  $4,923.47  $5,353.29
;
run;

Which procedure should I use to do this?  And can this procedure be done using "Tasks"?  I've seen PROC GLM in different blogs, but I fail to see why I would want to create a model with my data unless it outputs tests for whether my different samples have the same variance or not.  Any help is greatly appreciated; thanks!

-Valentine

[lvm · Accepted solution]

I don't know anything about Tasks in EG, so maybe someone else can answer. The most direct way really is to use PROC GLM. You would have to stack your data, adding a variable to identify the group, assuming you are trying to compare the variances of the three variables.  The HOVTEST option in GLM is easy to use (there are several sub-options). See:

https://documentation.sas.com/?docsetId=statug&docsetTarget=statug_glm_syntax15.htm&docsetVersion=15...

See example at:

https://documentation.sas.com/?docsetId=statug&docsetTarget=statug_glm_examples10.htm&docsetVersion=...

Regarding your concern about the need to fit a model: data analysis is almost always about fitting models, even when you don't realize it. Getting the mean and standard deviation for a single sample is equivalent to fitting a model of Y = mu + error. 

[davidvalentine]

This was very helpful information, Ivm, especially the part about stacking the data and adding a variable to identify the group(s); I never thought about doing this!

Thank you! 😄