Question about doing a loop
Posted: Thu Jul 15, 2010 3:24 am
Hi, I'm a newbie to eviews, here is my problem.
I want to a recursive modeling approach to decide which asset prices are good for predicting inflation rate. Suppose the time series running from 1980 to 2009 (quarterly data), asset prices are house price(a), stock index(b), interest rate(c), I use a rolling window of 30 observations, for each sub-sample(rolling window) I regress the inflation rate on different combinations of a, b and c, so 2 to the power of 3 of models will be estimated for each sub-sample, use AIC and BIC information criteria respectively to find the appropriate model for each sub-sample. So if there are 120 observations in the entire sample, I have to find the 90 models for the 90 sub-samples. Since I use separately AIC and BIC to choose models, there are be two groups of 90 models. Later I will do one-step ahead forecast and reply on squared forecast error to determine whether AIC or BIC is more accurate for forecasting.
[my objective is to find the right set of the 90 models and calculate the frequency of each asset prices appearing on right hand side of these 90 models, so the asset price appearing at the highest frequency will be most useful to predict inflation]
If anyone could help tackle this problem and point me to the right direction, I would really appreciate it.
I want to a recursive modeling approach to decide which asset prices are good for predicting inflation rate. Suppose the time series running from 1980 to 2009 (quarterly data), asset prices are house price(a), stock index(b), interest rate(c), I use a rolling window of 30 observations, for each sub-sample(rolling window) I regress the inflation rate on different combinations of a, b and c, so 2 to the power of 3 of models will be estimated for each sub-sample, use AIC and BIC information criteria respectively to find the appropriate model for each sub-sample. So if there are 120 observations in the entire sample, I have to find the 90 models for the 90 sub-samples. Since I use separately AIC and BIC to choose models, there are be two groups of 90 models. Later I will do one-step ahead forecast and reply on squared forecast error to determine whether AIC or BIC is more accurate for forecasting.
[my objective is to find the right set of the 90 models and calculate the frequency of each asset prices appearing on right hand side of these 90 models, so the asset price appearing at the highest frequency will be most useful to predict inflation]
If anyone could help tackle this problem and point me to the right direction, I would really appreciate it.