The RATS Software Forum

Hi Tom, I figured that the annual returns are a AR(1) process, and so your previous advice makes complete sense. So please ignore my first question. Could you please share any suggestions on the second question: dealing with a pair of variables when seasonality is high in one but not in another? Tha...

Hi Tom, Thanks so much for the advice. This helped and I was able to get to a converged model. Really appreciate your help! I have a couple of related questions. The annual returns that I use now (100*log(p{t}/p{t-12})), do not 'look' like a GARCH process. Please see attached. Do you think this is a...

Thanks a lot, Tom.
Please could you elaborate on why seasonal dummies might not be a good way to account for weather related seasonality?

Hi Tom, Yes these are log(p{t}-p{t-1}). They are prices of wholesale and retail maize in Zambia. Agricultural prices often show strong seasonality, more so in developing countries where markets are often incomplete and inter-temporal price arbitrage not as efficient in stabilizing seasonal price flu...

Hi Tom, Thank you for that very useful insight. I had also wondered if a VAR model would be more suitable to my analysis but could not find a solid reason behind it. This was very helpful. As for the ARMA mean model- it does not have to be the same for each variable and I am not doing it just to be ...

Hi Tom, Apologies for the late reply on this. Yes, I have been able to fit the univariate models for both. I am considering setting initial values with help of the estimates of the univariate model. Would you suggest that as a potential solution? Thanks so much for the help! Here are the codes and r...

Hi Tom, Thanks for the reply. I was exploring the simplest mean models that would provide no serial correlation in residuals for a reasonable number of lags (6-12 for my work). I wanted to keep the same mean model for both variables and ARMA(1,1) with monthly dummies appeared to be the best. I am pr...

Hi, I am trying to estimate a Bivariate GARCH in mean model. However, on running my code I receive warnings about non-convergence. I tried changing the NLPAR DERIVE option, but that does not help either. I do obtain some results that seem reasonable. However, I am worried that results of a procedure...