>For example in applying glm (general linear model)
>what should we do first ?
>do I look whether assumptions are satisfied ?
>or do I apply glm first, and after control which assumptions have
>problems ?
Hi,
I do not pretend to be anywhere in the league of experts that normally reply to
posts on this ng, but I always run descriptive and visual analyses first to see
if my data meet the required assumptions for the inferential tests I'm
interested in. Only after that do I either, run the test if all is okay,
possibly apply transformations or adjustments depending on the violation, or go
with a non-parametric alternative if required.
As for your inquiry regarding the availability of a flowchart, I get the
impression that many on this ng see no place for the use of flowcharts. Having
said this, as a graduate student who appreciated the same type of information
at one time, I highly recommend Table 2.1 in the Tabachnik and Fidell text
(Using Multivariate Statistics) entitled "Choosing Among Statistical
Techniques" It provides a decision-tree format that begins with consideration
of the type of research questions being asked and then supplements this with
information about the nature and number of dependent and independent variables
& the existence of covariates. Based on this information, an analytic strategy
is suggested along with the goals usually associated with these techniques.
Hope my post helps to illuminate some of your questions. Good luck.
Kathy L.
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
