[MATHEDCC] Logistic regression

Brian E. Smith (smithb@management.mcgill.ca)
Sat, 04 Jul 1998 23:48:26 -0400

Thanks to those who responded to my query on logistic regression. I am
embarrassed to say that I had an error in my data set (thanks John
Peterson) and the second ordered pair should read (12, 2100), not (2, 2100)
as originally reported. However the rest of the message is correct - and
adding the point (6, 2512) results in a domain error message.

Brian

>Date: Fri, 03 Jul 1998 12:57:08 -0400
>To: mcc
>From: "Brian E. Smith" <smithb@management.mcgill.ca>
>Subject: Logistic regression
>
>Does anyone have experience with logistic regression, particularly using
the TI-83. I am trying to fit some data points in a logistic regression
and the TI-83 works for some data sets but not others. For example, when I
run the regression on the data set
>
>(1, 3697)
>(2, 2100)
>(17, 1603)
>(20, 1127)
>
>I get the output: y = c/(1 + ae^(-bx)) with a = .4799, b = -.1028 and c =
5658.4
>
>However when I add one extra ordered pair (6, 2512) I get a domain error
message.
>
>Any idea what is going on?
>
>Also do you know of any other calculator or stats software that will do
the logistic regression (Minitab has a section called logistic regression
but it is something different - it seems to be an alternative to the GLM
regression model which is for use with binary, ordinal and nominal data sets.)
>
>Any help wil be appreciated.
>
>Thanks
>
>Brian
****************************************************************************
* To post to the list: email mathedcc@archives.math.utk.edu *
* To unsubscribe, send mail to: majordomo@archives.math.utk.edu *
* In the mail message, enter ONLY the words: unsubscribe mathedcc *
* Words in the Subject: line are NOT processed! *
* Archives at http://archives.math.utk.edu/hypermail/mathedcc/ *
****************************************************************************