Simple Regression

3 views (last 30 days)
Ramy Baly
Ramy Baly on 26 Feb 2011
Hello
I am new in Matlab. I have one predictor (X) with a response Y (binomial, 0s and 1s) for x, below a certain threshold => y=0 above the threshold => y=1
I want to make sure that linear regression can 100% predict that. how can I get the regression coefficients to prove that?
Regards Ramy
  1 Comment
John D'Errico
John D'Errico on 6 Mar 2011
I think perhaps you are new to statistics too. A statistical model (i.e., linear regression) will essentially never give you 100% certainty of prediction.

Sign in to comment.

Answers (1)

Manu R
Manu R on 6 Mar 2011
It sounds to me like you want to consider the logistic functional form for your linear regression, where:
Y(X) = a/(b + exp(c*X)), where a, b, c are fitting parameters.
If your response is binomial as you described, take a look at the multinomial logistic regression function MNRFIT in the MATLAB doc. Don't be confused by the examples on the doc page -- they're meant to illustrate true multinomial models where the response Y can have more than one dimension. This algorithm/function should work just as well for the problem you described.
You just want the basic options, i.e. use only the first two input arguments X and Y, make sure they're fit the dimensional requirements of the inputs and you should be set.
Hope this helps.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!