Can somebody please check my VaR model

Pm35 · July 12th, 2008, 9:09 am

I have attempted to develop a VaR model using Peaks over Threshold method (a branch of EVT)I have monthly hedge fund index returns and have computed the VaR but the difference at the 95 and 99 c.l seems negligible which has raised my concerns that i have done something wrong. I have then adopted the back testing approach -quadratic probability score (QPS) method as a comparison of alternative models as per (Lopez, (1998)).I have attached the model in excel maybe someone can tell me what i have done wrong.Cheers,Pierre

rks74us · July 25th, 2008, 6:18 pm

I took a quick look. Coupl eof questions:1. Not sure why are you restricting the number of data points to calculate the threshold? The "Small" formual only goes to row 139.2. You are using the same "Nu" value for P95 and P99.3. How did you calculate Beta and zeta?