wilmott.com

How to choose the dependent variable and set of independent variables if I regress S1 ~ S2+S3+S4Where S1,S2,S3,S4 are 4 different stock price.How to decide S1 should be dependent, or S2, S3, S4?

QuoteOriginally posted by: cygat6564This is an elementary question regarding the construction of a pairs trade. What is the correct way to calculate the weightings for each individual equity? Will a simple linear regression provide the correct weightings? Also does it matter if I am basing my trade off of percentage return over some period or the absolute price spread? -Thanks in advanceYou can take weigths proportional to average for example SimpleMovAverage of the equities. Also I saw articles where people use kalman filtration

In co integration relationship for pairs trading, we regress S1 ~ S2 without intercept. But we know, regression without intercept increases R square dramatically. How to handle this? R square change dramatically, how can we say that this fitting is not spurious?

Get yourself a copy of one of these books - is no generally accepted choice for how to determine the relative sizing of each side. I tend to use beta to scale them. This ensures that for a 1% change in the overall market, then both constituents of the pairs trade tend to have corresponding (expected) changes which are equal in value and opposite in direction.eg ABC $2.00 with beta 0.3and XYZ $5.00 with beta 1.5Consider the case that the market moves up 1% then we expect ABC to move up 0.3% and expect XYZ to move up 1.5%. Here we are using 'expect' in the mathematical sense. Using this method we might allocate $100k to ABC (ie buying 50k of them) and short $20k of XYZ (ie selling 4k of them). After this +1% change in the market then ABC will make $300 and XYZ will loose $300. In other words the expected return of the portfolio is zero. This is not the only way to do it but should be reasonably market neutral. Be aware that different people may measure beta using regression over different time frames meaning the number which I calculate for beta may be totally different to yours.Hope that helps.