Guys

Next week we're planning on recalibrating the dummified regression models ("placement models") built earlier this year and last for the Enron Credit pricing system. These basically have the objective of predicting a 5-yr CDS price for an illiquid name, based on available data. This data consists of:
Credit Ratings, Outlook and Watch
Equity based data - eg, EDF, 90d Volatility, Beta, Market Cap, Dividend yield, PE ratio
Financial ratios - eg, EBITDA interest coverage, Current ratio, EPS, Debt to Assets
Descriptive Info - Country and Industry

I've outlined the methodology used for the models thus far in the attached docs:

   
We're kicking off this effort next week with a brainstorming session with the trading desk, to outline areas of deficiency and proposed model improvements. From the quantitative side we'd appreciate greatly if any or all of you could read these docs and offer any suggestions for improvement of the models. Any ideas on statistical techniques or the overall methodology will be welcome. I'd also be happy to set up meetings or just talk on the phone if any further explanation is required.

Regards

Ben