The RATS Software Forum

Hi all,

I try to estimate a single equation 2-regime MS linear model using MsrRegression procedure. Does anyone know how I can extract the fitted values, the residuals, the raw and smoothed probabilities? In addition, how can I produce forecasts if I have forecast inputs and know the forecast states by modelling the regime one probability series with a logistic model? Lastly, I used the standard codes (example 9.1) in the e-course for the ML estimation, and I got the following error: MAT15, the subscripts are too large or non-positive error was evaluating entry 107.

The problem starts from here:

Set p1smooth = psmooth(t) (1)
Graph ....
Could anyone know how to fix this problem?

Many thanks

1. What do you mean by "raw" probabilities?
2. What do you mean by "residuals"?
3. What does "In addition, how can I produce forecasts if I have forecast inputs and know the forecast states by modelling the regime one probability series with a logistic model?" mean?

The problem with the PSMOOTH (isn't that the smoothed probabilities?) sounds like it's not defined over the whole workspace range. Restrict the SET instruction to the range of entries over which it's computed.

I would like to find the fitted values and the residuals of the model, and I do not know which functions to use. For example, for a garch model, one can save residuals to a predefined series by setting resids = name of the series. In addition, if I want to use the two-regime MS model to do forecasting, how can I do it using RATS?

Many thanks

See the article on page 3 of https://estima.com/newslett/October2013RATSLetter.pdf.