statistics question..please help
November 5, 2007 at 9:23 pm #48600
First of all, Thank you so much for taking the time to guide me.
I work for a financial services corporation and I am trying to analyze some data. I am trying to figure out the root causes that lead to foreclosure of a property. One of the main attributes that can lead to foreclosure is insufficient contact with the borrower. The more the number of contacts, the better it is.
I divided my data into 2 sets (one months worth of data).
The first set is the loans that did not end in foreclosure (Success). The second set is the loans that ended in foreclosure (Failures).
I have 5000 loans in first set and 1000 loans in the second set.
I did a Chi-square test to see if there is a statistically significant difference in the number of contacts between success and failures. The Chi-square test showed that there is a difference.
Now, my question is:
What is the tool that I can use that tells me the cut off point. What I mean is: How do I get a number X where I can say, if the number of contacts is less than X the probability of failure is higher.
Thanks0November 5, 2007 at 10:53 pm #164432
WardParticipant@pete Include @pete in your post and this person will
be notified via email.
Binary logistic regression can help answer the question. Minitab makes it pretty easy for this type of anlaysis/model. This thread has been moved to the Financial Services discussion forum. Please click here to continue the discussion.0November 6, 2007 at 1:45 am #164437
BrandonParticipant@Brandon Include @Brandon in your post and this person will
be notified via email.
Very interesting project – and about 50 years overdue. I started my career XX+ years ago in lending. I served some time in collections then origination. In origination, of course, we did credit check, income check etc. I always wondered if any of the data we collected and analyzed and based an approval on had anything to do with the probabilty of non-payment in the future.
I believe NOBODY in the lending industry ever tried to find a correlation between underwriting data and causal factors for default. Same approach year after year.
I love it – would like to hear the results.
However, before you dive too deep into the presumption that “staying in touch” is the Critical X – you should study your 1000 defaults for causal factors. Group them, pareto them and start studying any relationships you can identify. Then see if any of those causal factors could be assessed at underwriting – maybe and maybe not – things change that can’t be identified at underwriting. However, if you could identify some that could you could revolutionize how a loan is underwritten.0
The forum ‘General’ is closed to new topics and replies.