Solved
Contributor
Posts: 31

# how to interpret my graph?

Hi,

I used PROC GLM for ANOVA.

```ods html style=statistical;
ods graphics on;
title "Anova with Plots";
PROC GLM DATA = smoke3 order=data;;
CLASS amt;
MODEL fef=amt/solution;
CONTRAST 'ns vs ps' amt 1 -1 0 0 0 0;
CONTRAST 'ns vs ave inhale' amt 1 0 0 -.1 -.7 -.2;
CONTRAST '# of smo' amt 0 0 0 -14 -4 18;
CONTRAST 'ps vs ms' amt 0 -1 0 0 1 0;
ESTIMATE 'ns vs ave inh' amt 1 0 0 -.1 -.7 -.2;
ESTIMATE 'smo mn' intercept 70 amt 0 0 0 10 20 40 / divisor = 70;
ESTIMATE 'nonsmo mn vs smo mn' amt 35 35 0 -10 -20 -40 / divisor = 70;
ESTIMATE 'ps vs ms' intercept 70 amt 0 30 0 0 40 0 / divisor =70 ;
run;
quit;
ods graphics off;
ods html close;```

And i got following result:

Question: how to interpret my graph?

Thank you

Accepted Solutions
Solution
‎05-05-2016 09:48 AM
SAS Super FREQ
Posts: 3,907

## Re: how to interpret my graph?

See p. 3 of "Interpreting the Differences Among LSMEANS in Generalized Linear Models" by Robin High.

Briefly, the blue lines represent pairs of groups whose mean differences are significantly different from zero. (That is, the means are sig diff from each other.) The red lines represent pairs that are NOT significantly different from zero. In your example, the mean difference between the NI and PS groups are not sig diff. Similarly, the mean difference between the NI-LS groups and the LS-PS groups are not sig different. The other pairwise comparison of means are significantly different from each other.

All Replies
Solution
‎05-05-2016 09:48 AM
SAS Super FREQ
Posts: 3,907

## Re: how to interpret my graph?

See p. 3 of "Interpreting the Differences Among LSMEANS in Generalized Linear Models" by Robin High.

Briefly, the blue lines represent pairs of groups whose mean differences are significantly different from zero. (That is, the means are sig diff from each other.) The red lines represent pairs that are NOT significantly different from zero. In your example, the mean difference between the NI and PS groups are not sig diff. Similarly, the mean difference between the NI-LS groups and the LS-PS groups are not sig different. The other pairwise comparison of means are significantly different from each other.

🔒 This topic is solved and locked.