The type of regression to which you refer is called Deming Regression. See: https://en.wikipedia.org/wiki/Deming_regression
This should answer your question. http://ncss.wpengine.netdna-cdn.com/...Regression.pdf
Dear all,
I have been searching for a while without finding a solution, so I hope someone here can help.
Here is my problem:
I have (X, Y) data points. X is a roughness parameter for which I have several measurements, Y is the area of cells when they grow on a surface with the roughness X, for which I also have several measurements. Therefore for each pair (X;Y), X and Y are the mean values of a set of measurements, with standard deviations Sx for the X values and standard deviations Sy for the s values. One comment (I don't know if it matters): Sy does not correspond to an uncertainty in the measurement of a single cell, but to the variation of the area in a population of cells
To show the correlation between X and Y, I would like to perform a linear regression taking into account the standard deviations. I think I found a way to do that, using the approach described in the attached PDF.
First question: is that approach OK in my case ?
Second question : I would like to compute 90% confidence bands. Is there a simple way to do that ? I found the answer only for the case of a simple linear regression (with no standard deviations associated to X and Y measurements)
Thank your your help,
Quentin
The type of regression to which you refer is called Deming Regression. See: https://en.wikipedia.org/wiki/Deming_regression
This should answer your question. http://ncss.wpengine.netdna-cdn.com/...Regression.pdf
Thank you very much for your quick reply. The link for the PDF appears to be dead, could you attach it here ? Thanks again !
It worked great the first time, but I am getting the same error now. Here is a link to the cached version. http://webcache.googleusercontent.co...&ct=clnk&gl=us
Another resource: http://influentialpoints.com/Trainin...ssumptions.htm
Note: Deming Regression also goes by Orthogonal Regression and Errors-in-Variables regression.
Thank you very much, that helps a lot !
Tweet |