advanced method to compare regression coefficient

#1
Hello,

I am regressing two butterfly richness variables (summer and winter)
against a set of environmental variables separately.
Environmental variables are identitcal in each model.

In the summer model,
the weight rank of correlation coefficients is
temp > prec > ndvi.

The weight rank in winter is
temp > ndvi > prec.

As it is almost implausible to compare the coefficients directly,
pls advise any advanced method other than regression
to exhibit such trend on coefficient rank,

such as canonical correlation analysis (unsure if it is suitable here)

Thx


Elaine
 

bugman

Super Moderator
#2
I need a bit more info:

are you using this to make predictions? Or are you just looking at exploratory techniques at the moment?

CCA, may not be your best choice. Do you have some spatial component to your data? Or are your responses just considered from one site?
 
#3
comparing regression variables

I need a bit more info:

are you using this to make predictions? Or are you just looking at exploratory techniques at the moment?

=> now for exploratory techniques

CCA, may not be your best choice. Do you have some spatial component to your data? Or are your responses just considered from one site?
=> the response variable comprises of butterfly richness of 2000 grid (continuous distribution) spanning from 100 E to 130 E longitude, 18 to 25 N latitude.
 

bugman

Super Moderator
#4
CCA might be worth looking at in this case.

If you wish to see which variables discriminate between seasons, then you might want to try discriminant function analysis, which is kind of like a MANOVA in reverse.

As I'm still not entirely sure what you are trying to do, multiple regression or redundancy analysis might also be useful.
 
#5
advanced methods to compare regression coefficients

CCA might be worth looking at in this case.

If you wish to see which variables discriminate between seasons, then you might want to try discriminant function analysis, which is kind of like a MANOVA in reverse.

=> OK I will study discriminant function analysis at first.

As I'm still not entirely sure what you are trying to do, multiple regression or redundancy analysis might also be useful.
=> I did two multi-regression models and the weight rank difference was observed by variable coefficients in the models.

Please kindly indicate how multiple regression would be useful since the coefficients in distinct models could be compared directly.

Also, please kindly share why using redundancy analysis instead of canonical correspondence analysis in this case ? (How to judge whether the relationship between butterfly richness and environmental variables is linear or unimodal?)

Thank you.