Logistic regression vs. weights of evidence method

[E.Armijo]

Hi,

I noticed from some earlier publications (e.g., Soares-Filho et al
2002) that DINAMICA used to apply logistic regression (logReg) to
calculate spatial dynamic transition probabilities. The latest
versions use weights of evidence (WoE) method although I find it less
intuitive to explain to my colleagues and more difficult for inter-
variable comparison.
I wondered what considerations were behind this change in DINAMICA
development and what benefits can be attributed to the use of WoE
compared to the apparently simpler logReg. I am curious because I
have not been able to get any paper touching this.

regards,

E

[Britaldo Silveira Soares Filho]

The answer is simple: Spatial data often violate the assumptions of
statistical parametric methods, which is the case of Logistic regression.
So in principle, applying logistic regression method on map data, without
testing its basic assumptions, does not make sense - i.e. the results may
be biased. On the other hand, Weights of Evidence is a Bayesian method and
thus much more robust to deal with spatial data.

Anyhow, you could implement a logit equation in a calculate map using the
coeficients output from a logistic regression, as follows:

Pij = power(e, ax1 +bx2 +cx3 ...)/ (1 + power(e, ax1 +bx2 +cx3 ...))

Note that this equation is the same used in Weights of Evidence, where the
coefficents are replaced by the weights W+.

Best

Britaldo Silveira Soares Filho
Centro de Sensoriamento Remoto/UFMG
brit...@csr.ufmg.br
www.csr.ufmg.br
wwww.csr.ufmg.br/dinamica
553134095449, 34095417

> --
> You received this message because you are subscribed to the Google Groups
> "Dinamica EGO" group.
> To post to this group, send email to dinami...@googlegroups.com.
> To unsubscribe from this group, send email to
> dinamica-ego...@googlegroups.com.
> For more options, visit this group at
> http://groups.google.com/group/dinamica-ego?hl=en.
>
>

[Jean Francois Mas]

Logistic regression is based on the assumption that the relationship between
the transition probability and the explicative variable is a logistic
(sigmoid) function, which can be true or not.
As weights of evidence are calculated for each one of the category of the
explicative variable, the relationship between the transition probability and
the explicative variable is "function free" can very flexible.

Maybe weights of evidence approach does not have all the tools logistic
regression offers (forward and backward variable selection, statistics to test
significance, etc.) but is less sensitive to assumption violation with spatial
data. Moreover interpretation of the weights values (graphics) is very
intuitive and user is able to take into account expert knowledge. We carried
out a simple exercise of comparison between different approach, I send it to you

Regards

JF

On Mon, 7 Feb 2011 11:56:49 -0800 (PST), E.Armijo wrote

> --
> You received this message because you are subscribed to the Google
> Groups "Dinamica EGO" group. To post to this group, send email to
> dinami...@googlegroups.com. To unsubscribe from this group, send
> email to dinamica-ego...@googlegroups.com. For more options,
> visit this group at http://groups.google.com/group/dinamica-ego?hl=en.

*********************************************************
JF Mas
Centro de Investigaciones en Geograf�a Ambiental
Universidad Nacional Aut�noma de M�xico (UNAM)
Antigua Carretera a P�tzcuaro No. 8701
Col. Ex-Hacienda de San Jos� de La Huerta
C.P. 58190 Morelia Michoac�n M�XICO
tel. (443) 322 38 35
o (443)322 27 77 extensi�n 328 35
tel red UNAM (DF) (55)56 23 28 35
Fax Morelia (443) 322 38 80
Fax red UNAM (55)56 23 28 80

jf...@ciga.unam.mx
http://www.ciga.unam.mx/
******************************

[Eric Armijo]

Thanks for the replies. I will greatly appreciate a copy of the mentioned
article. Please send it to my e-mail address (rc...@hotmail.com).

regards,
E

-----Mensaje original-----
From: Jean Francois Mas
Sent: Tuesday, February 08, 2011 1:57 PM
To: dinami...@googlegroups.com
Subject: Re: [Dinamica EGO] Logistic regression vs. weights of evidence
method

[Ana Silva]

Hi,

Could you also send me the exercise created to compare logistic
regression with woe? (rrasterr(a)gmail.com)

Thanks in advance.

Regards,
Ana Silva

[gunawan tanjung]

Hi,

I am interesting with this comparison.
Could you also send me the comparison? (gunawan...@yahoo.com)

Thanks.

Best Regards,
Gunawan

 

---

From: Ana Silva <rras...@gmail.com>
To: dinami...@googlegroups.com
Sent: Thu, March 3, 2011 12:16:42 AM

Subject: Re: [Dinamica EGO] Logistic regression vs. weights of evidence method

>> email to dinamica-ego+unsub...@googlegroups.com. For more options,

>>  visit this group at http://groups.google.com/group/dinamica-ego?hl=en.
>
> *********************************************************
> JF Mas   

> Centro de Investigaciones en Geografía Ambiental
> Universidad Nacional Autónoma de México (UNAM)
> Antigua Carretera a Pátzcuaro No. 8701
> Col. Ex-Hacienda de San José de La Huerta
> C.P. 58190 Morelia Michoacán MÉXICO

> tel. (443) 322 38 35

> o (443)322 27 77 extensión 328 35

> tel red UNAM (DF) (55)56 23 28 35
> Fax Morelia (443) 322 38 80
> Fax red UNAM (55)56 23 28 80
>
> jf...@ciga.unam.mx
> http://www.ciga.unam.mx/
> ******************************
>

--
You received this message because you are subscribed to the Google Groups "Dinamica EGO" group.
To post to this group, send email to dinami...@googlegroups.com.

To unsubscribe from this group, send email to dinamica-ego+unsub...@googlegroups.com.

[Clément Parmentier]

Hi,

Could you also send me the exercise created to compare logistic

regression with woe? (clement.p...@mailoo.org)

Thanks in advance.

Clément

*********************************************************
JF Mas        
Centro de Investigaciones en Geograf�a Ambiental
Universidad Nacional Aut�noma de M�xico (UNAM)
Antigua Carretera a P�tzcuaro No. 8701
Col. Ex-Hacienda de San Jos� de La Huerta
C.P. 58190 Morelia Michoac�n M�XICO

tel. (443) 322 38 35

o (443)322 27 77 extensi�n 328 35

tel red UNAM (DF) (55)56 23 28 35
Fax Morelia (443) 322 38 80
Fax red UNAM (55)56 23 28 80

jf...@ciga.unam.mx
http://www.ciga.unam.mx/
******************************

[darlingt...@gmail.com]

Hi,

I am doing regression in excel using the logit add-in and the WoE as my input variable. When using WoE as your independent variable in logistic regression, what do you use as you dependent variable?

Thanks

Darlington 

[Oskr. C.R.]

Hi.

Could you please send me the exercise created to compare logistic regression with WoE?

Regards.