Hi all,I'm looking for some general help/advice regarding the backtesting of equity strategies. My question is (I think) quite basic, but I don't have a quant background and could use some help. I'm trying to backtest an equity strategy that ranks stocks on a combination of two factors (one is value-oriented and the other is momentum-related). First, I'll outline what I've done (which I believe to be incorrect): The output I'm working with is as follows: I have formed portfolios every other Friday from early 1993 to the end of 2006 (so each year has ~26 formations), each of which has a long side consisting of my top 50 ranked stocks on formation date and a short side with my 50 lowest ranking stocks on formation date. For each formation, I then pull 12-month, 6-month, 3-month, and 1-months returns for each stock. Next I take a straight arithmetic average of the long portfolio and the short portfolio, and I subtract short from long to arrive at portfolio return. This is all in Excel for now; the set up is as follows: Each column represents a portfolio formation date. Rows of interest include 12-month long, 12-month short, 12-month portfolio, 6-month long, 6-month short, 6-month portfolio, etc. For example, the column with header date September 24, 2004 will contain 12 rows corresponding to long, short, and portfolio returns based on formation date Sept 24, 2004 and holding periods to Sept 23, 2005, March 25, 2005, December 24, 2004, and October 22, 2004. The column to the left will contain the same data for a formation date of October 1, 2004. Concentrating on the 12-month holding period for now, I have taken the row for 12-month portfolio returns, subtracted the period-appropriate risk-free rate, and formed my y-variable. To test my strategy, I am (for now) interested in statistically significant alpha relative to the CAPM. As such, I have used Prais-Winston estimation with excess portfolio returns as my y-variable and excess market returns as my x-variable. I am taking the intercept as the alpha of my strategy relative to the CAPM and determining its statistical significance from my estimation output (which is done in Stata).Now, for my questions: First, is my choice of Prais-Winston appropriate? If not, what estimation method is most commonly used and/or most appropriate? Please keep in mind that my ultimate goal is to be able to state the alpha of the strategy relative to the CAPM and whether or not this alpha is statistically significant. I assume that once I've found the appropriate methodology, I can use it generally to test relative to other models such as Fama-French.Second, I am concerned about the fact that my data are overlapping. In practice, my portfolio will be held as follows: Implement long/short portfolio on formation date; hold for one year; implement portfolio again. This holding pattern makes me think I should use different return data for doing my backtest (e.g. form portfolio on Sept 24, 2004, find monthly returns of this portfolio for 12 months; form portfolio on Sept 23, 2005, find monthly returns for 12 months, etc; and then use this array of monthly returns to form my y-variable and regress against monthly market returns). My concern with changing my methodology is that I would lose ~25 portfolio formations per year. Thoughts on this?Thanks for any help with this; also, please feel free to offer additional suggestions/criticisms of my method.Thank you.

Apologies for the addendum, but I wanted to mention that if anyone could point me toward either papers or texts that address my question, that would be wonderful. I've searched the board and google extensively with no luck (perhaps I'm using a poor search string...any suggestions would be appreciated). Thanks for any help.

Hi There.I also want to apply Prais-Winston transformation. COuld you please help me out with this.RegardsGaurav