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

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

Highlighted
##

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

How do I interpret this linear regression?

Oct 8, 2020 8:14 PM
(223 views)

Hi everyone!

See the picture for the analysis I ran.

Simply put, I want to see if grant type (0, 1, 2, or 3) is predictive of total publication citations.

It looks like grant type is a significant predictor of citations. How do I report this?

Is something like *"Grant type was a significant predictor of total publication citations, with grant 2 having the most total citations compared to other grants"* appropriate?

Looking forward to hearing from you.

3 REPLIES 3

Highlighted
##

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Re: How do I interpret this linear regression?

Is this even the right analysis to run?

Highlighted
##

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Re: How do I interpret this linear regression?

Hi,

From the look of the report image captured in your first post, you may have a problem with the distribution of the residuals. The model assumes that residuals should be distributed normally: in your case it looks rather asymmetrical but I cannot be sure without actual testing. To do that, the model output, choose Save Residuals, select this new column in your table and test for normality.

From the look of it, you may want to try a modification of your your Total Citations with Log X or Log (X + 1) transformation.

Considering that you have a very simple model i.e. 1 Y by 1 X, and that the distribution of Ys may be problematic, you may also look into a nonparametric test (e.g. Wilcoxon) in the Fit Y by X > One Way Analysis platform > Nonparametric

These are my humble recommendations and input from more seasoned JMP users may be desirable.

Best,

TS

From the look of the report image captured in your first post, you may have a problem with the distribution of the residuals. The model assumes that residuals should be distributed normally: in your case it looks rather asymmetrical but I cannot be sure without actual testing. To do that, the model output, choose Save Residuals, select this new column in your table and test for normality.

From the look of it, you may want to try a modification of your your Total Citations with Log X or Log (X + 1) transformation.

Considering that you have a very simple model i.e. 1 Y by 1 X, and that the distribution of Ys may be problematic, you may also look into a nonparametric test (e.g. Wilcoxon) in the Fit Y by X > One Way Analysis platform > Nonparametric

These are my humble recommendations and input from more seasoned JMP users may be desirable.

Best,

TS

Thierry R. Sornasse

Highlighted
##

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Re: How do I interpret this linear regression?

If Grant Type 2 can only have values of 0, 1, 2, or 3, I would make it a Nominal variable rather than Continuous. An ANOVA approach would be more appropriate, I think. The analysis options would likely make more sense for you, too.

Dan Obermiller