Dear TalkStats user,

I am testing a trading strategy that I built using Matlab. I tested the strategy on the AEX (Dutch Stock Exchange) from 2000 until July 2012. The benchmark for my strategy is a regular buy&hold strategy.
My strategy returns 241.25% compounded return compared to a -55% compounded return on the benchmark.
My question concerns the data analysis of the returns. I am using Stata analyse the data and I use regressions to test if the constant remains positive after accounting for a momentum factor. I basically want to use the method of Jensen's Alpha.

my regression equation looks like this:

CAPM: R_i-R_f=α+ β(R_i- R_m )+ ϵ
2-factor model: R_i-R_f=α+ β(R_i- R_m )+β_(2 ) (MOM)+ϵ

Does this make sense?

My second question concerns autocorrelation. I tested the daily returns using the Durbin-Watson test and found autocorrelation to be present, as expected. I correct for autocorrelation using the Newey-West regression, i regress the excess returns of the trading strategy of the risk-free rate (3% annually) against the excess returns of the buy&hold strategy:

newey excessreturn_tradestrategy excessreturn_BuyHold, lag(N^0.25)

However this regression returns the same constant as a normal regression using

reg excessreturn_tradestrategy excessreturn_BuyHold:

I expected the constant to differ once i used Newey to correct for autoregression.

I also regressed using Prais-Winston regression:

prais excessreturn_tradestrategy excessreturn_BuyHold

However this regression returns a neqative return which seems strange to me because
I expected a positive constant since the trading strategy has much higher returns than the buy-hold strategy.

Can anybody help me with these issues and share his/her thoughts with me...

Thanks in advance,

Jelle