[Gretl-devel] Improvement of logit regression modelling in gretl

[andreas.rosenblad at ltv.se]

A few suggestions for improving the handling of logit regression models in
gretl:

1. Multinomial models, when the dependent variable is discrete but not
binary and does not have a natural ordering does not have a native
implementation in gretl yet. Please implement it. For example, include a
radio button in the GUI where the user can choose between using a
multinomial logit or an ordered logit (the current implementation) when the
dependent variable is discrete but not binary.

2. Currently, it is necessary that the lowest value for the dependent
variable is 0 and this is used as the reference category. Change this so
that any integer values can be used for the dependent variable and any of
these can be used as reference category.

3. Make it easy for the user to choose reference category by including
radio buttons for First category, Last category or Custom value, as in
e.g. SPSS. See the two attached screenshots.

Best regards
Andreas Rosenblad

(See attached file: logit suggestion.JPG)
(See attached file: SPSS logit.jpg)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: logit suggestion.JPG
Type: image/jpeg
Size: 40623 bytes
Desc: not available
Url : http://lists.wfu.edu/pipermail/gretl-devel/attachments/20080204/fd245fb4/logitsuggestion-0001.jpg
-------------- next part --------------
A non-text attachment was scrubbed...
Name: SPSS logit.jpg
Type: image/jpeg
Size: 18330 bytes
Desc: not available
Url : http://lists.wfu.edu/pipermail/gretl-devel/attachments/20080204/fd245fb4/SPSSlogit-0001.jpg