Dear list: Can I use simple linear regression when I have proportion data for both dependent and independent variables? Or, should I use beta regression analysis? Or any suggestion? Thanks! SH
simple linear regression with proportion data
5 messages · Steve Hong, Ben Bolker, S Ellison
SH <emptican <at> gmail.com> writes:
Dear list: Can I use simple linear regression when I have proportion data for both dependent and independent variables? Or, should I use beta regression analysis? Or any suggestion?
The distribution of the independent variable is irrelevant (in some circumstances it matters whether it is measured without error or not). Depending on what you want to do, and how close the proportion data come to 0 or 1, you might choose to use linear regression, or linear regression on arcsine-square-root transformed data, or beta regression. It really depends what you want to do with the answers and what your audience expects. You might try this on http://stats.stackexchange.com with a bit more context. Ben Bolker
Very useful comment and helpful website! Many thanks to you!!! SH
On Fri, Nov 16, 2012 at 5:16 PM, Ben Bolker <bbolker at gmail.com> wrote:
SH <emptican <at> gmail.com> writes:
Dear list: Can I use simple linear regression when I have proportion data for both dependent and independent variables? Or, should I use beta regression analysis? Or any suggestion?
The distribution of the independent variable is irrelevant (in some circumstances it matters whether it is measured without error or not). Depending on what you want to do, and how close the proportion data come to 0 or 1, you might choose to use linear regression, or linear regression on arcsine-square-root transformed data, or beta regression. It really depends what you want to do with the answers and what your audience expects. You might try this on http://stats.stackexchange.com with a bit more context. Ben Bolker
______________________________________________ R-help at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-help PLEASE do read the posting guide http://www.R-project.org/posting-guide.html and provide commented, minimal, self-contained, reproducible code.
2 days later
-----Original Message-----
Can I use simple linear regression when I have proportion data for both dependent and independent variables? Or, should I use beta regression analysis? Or any suggestion?
The distribution of the independent variable is irrelevant (in some circumstances it matters whether it is measured without error or not).
Agreed that if you just want a line that goes somewhere near the data you can do pretty much anything.
But don't those circumstances you referred to include 'any time you want an unbiased estimate of the slope or a reliable standard error on coefficients'?
S
*******************************************************************
This email and any attachments are confidential. Any use...{{dropped:8}}
On 12-11-19 10:18 AM, S Ellison wrote:
-----Original Message-----
Can I use simple linear regression when I have proportion data for both dependent and independent variables? Or, should I use beta regression analysis? Or any suggestion?
The distribution of the independent variable is irrelevant (in some circumstances it matters whether it is measured without error or not).
Agreed that if you just want a line that goes somewhere near the data you can do pretty much anything. But don't those circumstances you referred to include 'any time you want an unbiased estimate of the slope or a reliable standard error on coefficients'?
Did you see that I wrote "distribution of the **independent** variable" above? Ben