real world regression models

Hi guys,

after reading a lot of topics about multiple regression models I’m curious to take a look into real model (most probably, Excel file) made for practical purpose and reflecting real industry’s calculations.

If any of you can share such models with me it would be very appreciated)

Just run one on your own. All of the stuff I have is proprietary to my firm and would bet most people would be in the same boat. Just take something like GDP and try to explain it with BBB spreads, unemployment, or other variables you choose and see how it turns out. You could also do equity market ones if that is more interesting to you.

Alternatively you could read the fama-french paper. On the dimensional funds website I believe there is a paper on there for one of their fixed income models.

Add data analysis toolpack into your excel its easy to run a regression. Download FRED plugin and you can get economic data from the Fed. If you used the CFAI material the format of the output will look familiar, they used excel to do regression

Thanks for replies, guys.

I even wrote a request to BIWS, and the guy was so kind to reply, answering that IB industry doesn’t used a lot regression analysis, and most of it is pretty simple.

I’m familiar with Data Analysis add-in, but i would prefer to take a look into real example. For instance, in mining industry this tool can be very helpful in order to forecast commodities’ prices. I think gold prices can be regressed against China/India population, M1,M2 money supply in USA, gold production and exploration results, etc. Than execute all these analysis regarding to heteroskedasticity, autocorrelation, multycollinearity, misspecifications, etc. It sounds a way academic, but i wasn’t able to find any examples on Internet.

once you get into that level of regression you should certainly be using R or something like that if you have not already. You would be able to do something like that quite simply in statistical softwhere once you know the language. My gf is in biostatistics and uses SAS but the concept is the same and could whip up something like that rather quickly. From what I have read on many places, R is basically what anyone who does statistical analysis does.

As far as your BIWS question I would imagine he is correct, an IB analyst isn’t really in a place to be doing that. Im sure more ER guys, and certainly quant guys are using statistical modeling.

Excel isn’t really good to use, especially without the right plug-ins (more tedious to be thorough). SAS is the gold standard for statistical analysis (at least, according to people in graduate/PhD stats programs), and R, SPSS, Minitab, and Stata are all popular as well. R is free, so it has gained traction, but some journals won’t accept output from it because it’s open-source (at least, I’ve heard that).

You can perform at least some of those things (that you mentioned) in Excel, but it’s much more cumbersome than a statistical software package.

Good point regarding the ease with which another software can do this for you! If you’re doing any kind of regression, you really should be checking the error assumptions regarding specification (mean error of zero), normality, homoscedasticity, and no autocorrelation (or considering if autocorrelation is a concern for your data type).

Check out any of the IMF’s World Economic Report. There’s usually 2-3 studies per report and plenty of references to previous studies.

Thanks, guys!