Re: Logistic regression and linear regression
- From: Richard Ulrich <Rich.Ulrich@xxxxxxxxxxx>
- Date: Mon, 26 May 2008 17:02:30 -0400
On Mon, 26 May 2008 06:44:12 -0700 (PDT), maria.hartberg@xxxxxxxxx
wrote:
Hello,
Would I get the same results if I turned my binary data into logratios
and then run a linear regression rather than running a logistic
regression?
No.
For grouped data with no 0/100s, a Least-squares
regression program *can* be used, with weighting,
to get a "Least chi-squared" solution. Finney gave
details, in his book (about 1972) on bioassay, if not
the logistic book. You might be able to find me citing
that years ago, if you Google-groups
<group:sci.stat.* author:ulrich Finney >.
The reason for this is that I need to run stepwiste forward logistic
regression in matlab but matlab has no function for that so I thought
that maybe I could turn my binary variables to log ratios and then do
a linear.
PS. Maybe the question seems simple but I am not a statistics person.
Stepwise solutions are usually not a good idea. You
could also Google-groups for me on "stepwise".
--
Rich Ulrich
http://www.pitt.edu/~wpilib/index.html
.
- References:
- Logistic regression and linear regression
- From: maria . hartberg
- Logistic regression and linear regression
- Prev by Date: Re: Comparing Subgroup Percentiles
- Next by Date: Re: TESTING HYPOTHESES BY MONTE CARLO
- Previous by thread: Logistic regression and linear regression
- Next by thread: WORMTOMSKY finding to prove a coin is FAIR
- Index(es):
Relevant Pages
|
