Search
Question: What do ellipses in 'plotarrays' function represent (from made4)
0
gravatar for maartenvandijk
2.5 years ago by
European Union
maartenvandijk0 wrote:

Hello all, 

I'm plotting the results of a correspondence analysis using the "plotarrays" function in made4, one of the arguments is the "ellipse" for which the default input is '1', my question is: what do the circles represent? 1 x standard deviation? 95% CI? 

Thanks & regards, Maarten. 

ADD COMMENTlink modified 2.1 years ago by Aedin Culhane510 • written 2.5 years ago by maartenvandijk0
0
gravatar for Aedin Culhane
2.1 years ago by
Aedin Culhane510
United States
Aedin Culhane510 wrote:

Hi Maarten

Sorry for the long wait for a reply, the ellipse is simply an inertia ellipse. Its just meant to highlight the individuals of the group and shows most of the group were projected within that location on the axes. It is not a confidence ellipse. To turn off the ellipse, set ellipse to 0.  To draw a bigger or smaller ellipse change it to 2.5, 0.5, whatever you wish. A larger cellipse will draw a larger circle around the points in the graph.  The ellipse just summaries the graph and does not represent any confidence. 

plotarrays, calls plotting functions in the R package ade4. Ellipses are drawn by the function scatterutil.ellipse.  Have a look at the code, its pretty simple.

plotarrays gives the size of the ellipse 1.0 to the paramater cellipse in the ade4 function scatterutil.ellipse. The default cellipse is 1.5.

 

In plotarrays, I changed it from cellipse from 1.5 to 1.0, I probably found a smaller ellipse to be more useful given the noisy data I was analyzing. Feel free to change it back to 1.5.  

This has been much discussion about this on the R forum http://lists.r-forge.r-project.org/pipermail/adegenet-forum/2011-May/000281.html which links out to a more detailed discussion the ade4 forum  http://listes.univ-lyon1.fr/wws/arc/adelist/2011-05/msg00003.html (you need to click on the button "I am not a spammer" and it will take you to the discussion).  If you can read french, also see  http://pbil.univ-lyon1.fr/R/pdf/qr3.pdf, however feel free to ignore the section in this pdf document where they discuss the groups are samples drawn from a bivariate  normal distribution, this is unlikely to hold for most data.

There are some recent graph packages, I like ggord which draws pretty plots but is a little limited. Its available on github. 

Best 

Aedin

ADD COMMENTlink written 2.1 years ago by Aedin Culhane510
0
gravatar for Aedin Culhane
2.1 years ago by
Aedin Culhane510
United States
Aedin Culhane510 wrote:

Hi Maarten

Sorry for the long wait for a reply, the ellipse is simply an inertia ellipse. Its just meant to highlight the individuals of the group and shows most of the group were projected within that location on the axes. It is not a confidence ellipse. To turn off the ellipse, set ellipse to 0.  To draw a bigger or smaller ellipse change it to 2.5, 0.5, whatever you wish. A larger cellipse will draw a larger circle around the points in the graph.  The ellipse just summaries the graph and does not represent any confidence. 

plotarrays, calls plotting functions in the R package ade4. Ellipses are drawn by the function scatterutil.ellipse.  Have a look at the code, its pretty simple.

plotarrays gives the size of the ellipse 1.0 to the paramater cellipse in the ade4 function scatterutil.ellipse. The default cellipse is 1.5.

 

In plotarrays, I changed it from cellipse from 1.5 to 1.0, I probably found a smaller ellipse to be more useful given the noisy data I was analyzing. Feel free to change it back to 1.5.  

This has been much discussion about this on the R forum http://lists.r-forge.r-project.org/pipermail/adegenet-forum/2011-May/000281.html which links out to a more detailed discussion the ade4 forum  http://listes.univ-lyon1.fr/wws/arc/adelist/2011-05/msg00003.html (you need to click on the button "I am not a spammer" and it will take you to the discussion).  If you can read french, also see  http://pbil.univ-lyon1.fr/R/pdf/qr3.pdf, however feel free to ignore the section in this pdf document where they discuss the groups are samples drawn from a bivariate  normal distribution, this is unlikely to hold for most data.

There are some recent graph packages, I like ggord which draws pretty plots but is a little limited. Its available on github. 

Best 

Aedin

ADD COMMENTlink written 2.1 years ago by Aedin Culhane510
0
gravatar for Aedin Culhane
2.1 years ago by
Aedin Culhane510
United States
Aedin Culhane510 wrote:

Oops sorry for posting twice. I'm not sure what I did wrong. 

ADD COMMENTlink written 2.1 years ago by Aedin Culhane510
Please log in to add an answer.

Help
Access

Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
Powered by Biostar version 2.2.0
Traffic: 200 users visited in the last hour