topic Re: One way ANOVA outputs in SAS in Statistical Procedures

One way ANOVA outputs in SAS

ACLAN — Thu, 29 Apr 2021 12:45:49 GMT

Hi all,

I'm running a one way ANOVA analysis in SAS based on the following code:

proc anova data=dt;
class SERINDEX;
model Value0USNumdec = SERINDEX;
run;

The output returns:

The SAS System

The ANOVA Procedure

Dependent Variable: Value0USNumdec

Source DF Sum of Squares Mean Square F Value Pr > FModelErrorCorrected Total

4	0.00220000	0.00055000	27.50	<.0001
20	0.00040000	0.00002000
24	0.00260000

R-Square Coeff Var Root MSE Value0USNumdec Mean

0.846154

0.912681

0.004472

0.490000

Source DF Anova SS Mean Square F Value Pr > FSERINDEX

0.00220000

0.00055000

27.50

<.0001

As an R user, I'm used to get the DF, Variance component (VC), CV (as percentage), SD and Root MSE both for the model and for the error when I perform an one-way ANOVA analysis.

How can I specify this output in SAS? The user guide for Proc Anova has nothing on it. I also looked at the proc Varcomp or proc mixed procedures but they seem not to give the output I need as well.

Re: One way ANOVA outputs in SAS

PaigeMiller — Thu, 29 Apr 2021 14:01:27 GMT

You really ought to format your output so that it is readable, that will help us and thus you get better answers and quicker answers. For output, a screen capture is fine (and readable).

I'm used to get the DF, Variance component (VC), CV (as percentage), SD and Root MSE both for the model and for the error

I do not know what the "variance component (VC)" in a one-way ANOVA means. Can you explain/provide a formula?

The SD**2 is the "Mean Square" column. There is no such thing as a root MSE both for the model and for the error. It only exists for the error, and is shown. The DF and CV are shown.

Re: One way ANOVA outputs in SAS

ACLAN — Thu, 29 Apr 2021 14:10:49 GMT

This is the R output I'm used to where the "estimates" for the errors and the model are shown. I wanted to create a similar output in SAS where, for both the error and the model I have the DF, SD and CV (%). Is there any option or another proc statement that I can use to obtained something similar to the image above?

Re: One way ANOVA outputs in SAS

PaigeMiller — Thu, 29 Apr 2021 14:25:06 GMT

I asked you to explain what VC means in a one-way ANOVA, or provide the formula, I don't see that.

Everything else you ask for is in the ANOVA output, or can be computed from the ANOVA output.

There is nothing in SAS that produces this exact output; but you can of course program something to produce this exact output (although I am skeptical that matching the R output appearance exactly is necessary, or worth the effort)

Re: One way ANOVA outputs in SAS

SteveDenham — Fri, 30 Apr 2021 16:55:49 GMT

Looking at this output, i notice a line for total, one for the one-way factor and one for the error. The total variance is that obtained from a null model (no effects). The error variance is the residual error after fitting the model, and the attributed variance component for the one-way factor is the difference between those two. Now comes the tricky part to think about, and it comes from the usual definition of variance component - the variance that is attributable to some random effect. Now if you assume that SERINDEX is a random effect (and thus that this is a simplified nested model), then that value in the R output is then a random variance component.

But in PROC ANOVA, SERINDEX would be fit as a fixed, non-nested effect. Perhaps PROC NESTED would give you what you want. Without the dataset it is hard to tell.

SteveDenham