Regression Analysis - Needed for research project

[Ecomath]

Hello,

I'm a second year mathematics and economics student, and I've been hired by an economic development organisation to conduct a research project on the probability of loan default in micro-credit borrowers in rural Kenya (I'll be heading there in person this summer).

Basically, I'll be combing through a partner micro-finance institution's data on previous loan recipients in order to identify factors that might indicate a heightened risk of default. For example, whether female clients are more reliable than male clients; or perhaps whether clients with more education are safer than those with less.

Since I'm quite early in my degree, I haven't had a formal introduction to regression analysis. I understand the basic concepts of probability and variance, but I'm seriously lacking in detail.

I expect them to train me on the statistics needed, but I'm looking to get a head start. For a project such as this, what tools will I need to analyse the data appropriately (to identify and then test the strength of these relationships)? And are there any books you'd recommend?

Cheers,
Ecomath

 

[Stephen Tashi]

Unless you know that you're expected to use regression analysis, don't focus on that subject. It may be that you'll be expected to use particular software to do the job. Find out if your employers like particular software tools. That may lead to them telling you what techniques they do with the tools.

If your employers don't really know really know what they're doing and they expect you to be a trail blazer, I suggest you look up articles on analyzing the risk of loan defaults on the web. (I think it's a popular topic.) When you see what math is in those articles you can ask the forum about it.

 

[chiro]

Hey EcoMath and welcome to the forums.

Your task is not an easy one and I would exercise caution about doing that kind of work if you are unfamiliar with statistics and probability in a deep enough manner.

If you are using specific software that implements specific techniques then that is one thing, but if you trying to statistical analyses that are the result of a dynamic problem where you are going from first principles then this requires you to have a proper statistical mindset and this comes from a lot of experience.

My point is to exercise caution and be aware of not only what you know, but what you don't and if you don't know something then don't get involved (or get someone who does that you can validate).

 

[Ecomath]

Unless you know that you're expected to use regression analysis, don't focus on that subject. It may be that you'll be expected to use particular software to do the job. Find out if your employers like particular software tools. That may lead to them telling you what techniques they do with the tools.

If your employers don't really know really know what they're doing and they expect you to be a trail blazer, I suggest you look up articles on analyzing the risk of loan defaults on the web. (I think it's a popular topic.) When you see what math is in those articles you can ask the forum about it.

Thanks for the advice. It was my employer who suggested that I learn regression analysis, and it's my understanding that they'll be providing some training.

There will be software tools that I can use for calculation, but I'd like to understand the theory in order to better analyse the data.

I've picked up Regression Analysis: Theory, Methods and Applications. It's a graduate-level book, but I've been understanding it quite well, and it seems pretty relevant.

I was just looking for confirmation that I'm on the right track.

 

[Stephen Tashi]

I've picked up Regression Analysis: Theory, Methods and Applications. It's a graduate-level book, but I've been understanding it quite well, and it seems pretty relevant.

I'm not familiar with the book. Your original question uses the illustration of a discrete variable ( "male" vs "female"). The garden variety of regression analysis doesn't deal with such variables, but there are specialized types that do. If your example is apt, does that book cover this specialized type of regression?

Is it possible that your employer is using the term "regression analysis" to include the technique called "analysis of variance" (ANOVA)?