Turn on suggestions

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

Showing results for

- JMP User Community
- :
- Discussions
- :
- correlation

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
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

correlation

Apr 20, 2018 8:13 PM
(2054 views)

Hi

Is this true that in some analyses, especially Microbial diversity, linear correlations are weak, but still biologically significant by weak I mean 0.27 or -0.14, 0.07,-0.08? I dont agree that this true

if a correlation is weak but significant would it mean it is correlated having above values.

Any comments?

Thanks

7 REPLIES 7

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

Re: correlation

Statistical significance of a correlation does not mean that the correlation is relivant or important. Statistical significance basically implies that the value of the correlation is probably a true relationship. And even correlations of .27 or -.14 can be significant, but the amount of the relationship between the measurements isn't very strong.

Jim

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

Re: correlation

Thanks Jim

This is what I thought. knowing that the number is very small would you count this as a weak correlation or would you neglect it ?

This is what I thought. knowing that the number is very small would you count this as a weak correlation or would you neglect it ?

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

Re: correlation

There is not a fixed rule on this. The topic of study has to be taken into account.

Jim

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

Re: correlation

Thank you Jim

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

Re: correlation

Jim provided an excellent response. This is just a little added advice: make sure you are checking your data/model for non-linearity, outliers, unaccounted variables of time, temperature or mixtures.

Open this link from University of Florida Health and scroll down to the plot of Fuel Used vs. Speed, it is a picture of a strong relationship, but a small correlation negative linear correlation.

http://bolt.mph.ufl.edu/6050-6052/unit-1/case-q-q/linear-relationships/

Of course a classic example can be found in the Anscombe data. Run the script below. Y1 is what is expected for a 0.66 RSquare ( approx 0.816 linear correlation). Y2 and Y3 have a much stronger relationships: Y3 is weakened by an outlier and Y2 has a quadratic relationship with X (wrong model). Then look at Y4, the data results look more like an outlier, the result is inconclusive. Diagnostic plots and residual anayses are recommended procedures for and "modeling" of relationships.

```
dt = open("$sample_data/Anscombe.jmp");
dt << run script("The Quartet");
qwin = Window("The Quartet");
xx= (qwin<<child) << Xpath("//OutlineBox[@helpKey='Bivariate Report']") ;
yy = xx << get scriptable object;
yy << Show Points;
```

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

Re: correlation

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

Re: correlation

Thanks