How to read Ljung Box Test and what to do with the results?
Posted: Sun Jul 10, 2011 12:20 pm
Hello forum,
I have been reading a lot of helpful things on this forum so far and wanna say thank you first of all to all the active people here.
Unfortunately I couldnt find a solution to my problem yet: I have daily returns for several stock market indices and want to test for market efficiency and integration with a LS regression model. For the residuals I want to use an asymetric GARCH model. When describing the data, I want to tell the reader also about autocorrelation. I applied to Ljung-Box Test in levels (as I already have daily returns). But what can I do with the output? For lag 1 and 2 the results are mixed - half of the sample has no autocorrelation, the othr half does (assuming I m reading Prob. column correctly: Prob. of 0.0000 rejects the Null Hypoth of No Autocorrelation, right? - so there is AC). After 10 lags, all series have P = 0.00001, such that my conclusion is AC for all series. Am I interpreting this correctly? What does it mean if I have No AC in the first 2 lags, and AC after 3 lags onwards? Do I have to change my regression model(where I use 1 lag all the time)?
My regression model is specified as R_Index_B(0) = c + R_Index_B(-1) + R_Index_A(-1) + Residual_Index_A(-1)
where residual_index_A is defined as R_Index_A(0) - R_Index_A(-1)
Cheers,
Dima
I have been reading a lot of helpful things on this forum so far and wanna say thank you first of all to all the active people here.
Unfortunately I couldnt find a solution to my problem yet: I have daily returns for several stock market indices and want to test for market efficiency and integration with a LS regression model. For the residuals I want to use an asymetric GARCH model. When describing the data, I want to tell the reader also about autocorrelation. I applied to Ljung-Box Test in levels (as I already have daily returns). But what can I do with the output? For lag 1 and 2 the results are mixed - half of the sample has no autocorrelation, the othr half does (assuming I m reading Prob. column correctly: Prob. of 0.0000 rejects the Null Hypoth of No Autocorrelation, right? - so there is AC). After 10 lags, all series have P = 0.00001, such that my conclusion is AC for all series. Am I interpreting this correctly? What does it mean if I have No AC in the first 2 lags, and AC after 3 lags onwards? Do I have to change my regression model(where I use 1 lag all the time)?
My regression model is specified as R_Index_B(0) = c + R_Index_B(-1) + R_Index_A(-1) + Residual_Index_A(-1)
where residual_index_A is defined as R_Index_A(0) - R_Index_A(-1)
Cheers,
Dima