turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- How to interpret GLM ouput?

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

01-28-2015 08:21 AM

I'm new user to Modelling and now I've struck whilst interpreting the GLM output.

What does F value and Pr > F denotes? I've F value as 49.63 Pr > F as <.0001? When I should reject my null hypothesis for 95% CI and what is the significance behind F value?

Thanks for any help you offer.

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

01-28-2015 08:47 AM

The F-statistic compares two variances by dividing them, in the simple case of a one-way PROC GLM model, it would be the variance* of the model divided by the variance* of the residuals. A large number for the F-statistic indicates that the variability of the model term being tested is larger than the error variability, and if this number is large enough, then you say the F-statistic is statistically significant (in layman's terms, the F-statistic wouldn't get this large by random chance, and so we conclude that the result is due to a statistically significant effect). The Pr>F idnicates the probability of getting this large of an F-statistic if the null hypothesis was true. A Pr>F of less than 0.05 would indicate the effect is statistically significant with 95% confidence.

All of this is basic hypothesis testing, and if my paragraph above is not clear to you, then you probably ought to read a text on basic statistics and hypothesis testing.

* — well actually these are not variances but rather mean squares from the model, but they are very similar to variances

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

01-29-2015 03:09 AM

What if the F value is more than 1000 ,e.g.2000. Still would be significant?

How the critical value being calculated in F-statistic? Whether the critical value keeps changes when degrees of freedom changes.?

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

01-29-2015 04:00 AM

"What if the F value is more than 1000 ,e.g.2000. Still would be significant?"

Yes. I think so. since all the statistical estimator is estimating a deviation. 1000 is the F value when H0 is true. F=1000 is already significant , so F=2000 would be more significant (i.e. the deviation is bigger).

"

How the critical value being calculated in F-statistic?"

F estimator is a combo of two Normal distributions , check its formula at wikipedia.com

"Whether the critical value keeps changes when degrees of freedom changes.?"

No. I don't think so . It is depend on your significant degree ALPHA . i.e. is base on the ALPHA percentile .

It have nothing to do with DF .

Xia Keshan

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

01-29-2015 04:40 AM

Thanks for your inputs.

'1000 is the F value when H0 is true' - if F value is significant then we should rejecting the null hypothesis. isn't? So it means H0 is NOT true. Apologize if I'm wrong.

'P' value will be 0.05 for 95% CI. Do we've any standard (like P value) F critical value for 95% CI?

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

01-29-2015 05:09 AM

Yes. You are right .

F critical value for 95% CI is the 95% percentile of F distribution.

P=0.05= Probability( F > F0)