I am currently working on my Master Thesis in Finance. I am researching determinants of Credit default Swap Spreads, by executing linear regressions on my time series data. No my supervisor asked me to test for structural breaks. I've tested if in the whole model, any of my regression coefficents breaks, and I've tested each regression coefficent independently for breaks. I used Chow-testing, which gave breaks in many weeks for the whole model, and also many breaks in testing each coefficient independently. Following I've used the QLR ratio, and tested my trimmed F-statistic (15%) against Quandt's critical values, which resulted in less breaks, but still to much to draw conclusions from. Then my supervisor suggested to use a likelihood ratio test, which I've been trying. But the syntax i've found on the matter uses logit regressions, and since I dont have a binary dependent, it does not work.
I want to test my if my model changes, but I am running out of ideas here. Anyone any thoughts?
Advertise on Talk Stats