R help - Oct 2008 - Re : (a) Credit Scoring models and (b) aceesing earlier emails

Un texte encapsul? et encod? dans un jeu de caract?res inconnu a ?t? nettoy?...
Nom : non disponible
URL :
<https://stat.ethz.ch/pipermail/r-help/attachments/20081014/df8fd0b2/attachment.pl>

Thanks Mr Justin,
 
The site indeed contains lots of articles. But my immediate query is What is the
process of forming the credit scoring model i.e. how do you construct the credit
scoring card and what are the steps one should follow?
 
As I need to meet the deadline, I need this information on top priority basis.
 

Saggak
 

--- On Tue, 14/10/08, justin bem <justin_bem@yahoo.fr> wrote:

From: justin bem <justin_bem@yahoo.fr>
Subject: Re : [R] (a) Credit Scoring models and (b) aceesing earlier emails
To: saggak1908@yahoo.co.in
Cc: "R Maillist" <r-help@stat.math.ethz.ch>
Date: Tuesday, 14 October, 2008, 3:04 PM





Go to this link http://defaultrisk.com/ you will find a lot about credit scoring


 Justin BEM
BP 1917 Yaoundé
Tél (237) 99597295
(237) 22040246




----- Message d'origine ----
De : saggak <saggak1908@yahoo.co.in>
À : r-help@r-project.org
Envoyé le : Mardi, 14 Octobre 2008, 9h11mn 53s
Objet : [R] (a) Credit Scoring models and (b) aceesing earlier emails

Hi!
 
I have been reading lots of queries regarding logistic regression as well as
Credit scoring model and the related replies, which I must admit are full of
wisdom and help us to understand where do we stand. The contributions of Frank E
Harrell Jr, etc are invaluable. Incidentally I am also trying to work on Credit
scoring model and initially I tried to use the multiple regression etc. and now
I am using logistic regression using R.
 
I have two queries.
 
(a)            Can someone guide me as to how the credit score model should be
developed step by step? I mean what are the steps involved in developing credit
scorecard model? I have a sample database of 78000 bank customers and the period
is April 2004 – March 2007 and I also have the their default / non – default
position as on March 01, 2008. (There are 13756 defaulters). So using these
defaulters (coded as 1)  / non defaulters (coded as 0) and using logistic
regression, I have arrived at regression coefficients and then the Probability
of default for each of these 78000 customers. Can someone guide me regarding the
flow of credit scoring model? i.e. What are the steps involved in developing
Credit scoring models? I don’t have access to the books, as these are most of
the times very expensive and my organization being small and new, can’t afford
buying them as of now. I hope you understand my problem.
 
(b)            If I want to refer to earlier mails (I have recently become
member of this R group), how do I access these mails e.g. suppose I want to
refer to all the mails dealing with logistic regression, how do I do it?
 
Saggak


      Connect with friends all over the world. Get Yahoo! In
    [[alternative HTML version deleted]]





      Add more friends to your messenger and enjoy! Go to
http://messenger.yahoo.com/invite/
	[[alternative HTML version deleted]]

> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-
> project.org] On Behalf Of saggak
> Sent: Tuesday, October 14, 2008 4:20 AM
> To: justin bem
> Cc: r-help at r-project.org
> Subject: Re: [R] Re : (a) Credit Scoring models and (b) aceesing
> earlier emails
>
> Thanks Mr Justin,
>
> The site indeed contains lots of articles. But my immediate query is
> What is the process of forming the credit scoring model i.e. how do you
> construct the credit scoring card and what are the steps one should
> follow?
>
> As I need to meet the deadline, I need this information on top priority
> basis.
Then I would suggest that you hire someone with expertise in this area to do it
for you or at least help you with it.

I am far from being an expert in credit score models, but the look of your
original question suggests that the answer to your question would be a short
course in the area or part of a full college course.  Even if someone on the
list was willing and able to give you a full answer (and no one is obligated
to), I doubt that a full answer could be given in this format.

It is a mistake in any statistical procedure to assume that a simple set of
steps can give a complete answer without a full understanding of why those steps
were taken, why others were not, and when to do which steps.

You have been pointed to some resources, and there may be others available.  If
you want to do any type of modeling right, then you need to be willing to invest
the time and effort needed into understanding the full topic.  If you are not
willing/able to invest that time and effort, then hire someone who has invested
the time and effort to understand.


--
Gregory (Greg) L. Snow Ph.D.
Statistical Data Center
Intermountain Healthcare
greg.snow at imail.org
801.408.8111