Dear all,
I am a new statalist user.
I think I want to use arima model but I am unsure if it is ok for interrupted time series with deterministic trend and I have some difficulties in interpreting coefficients. I am using stata10 for windows.
I have monthly time series with deterministic trend. Time series is trend stationary. For my research, I am interested in changes in time trend after intervention. I know it can be done with Prais-Wintsen regression but it works only for AR(1) code: prais depvar time intervention time_after_intervention
I modelled ARIMA(2,0,3) process with time, intervention, timeafter intervention as independent variables (arima depvar time intervention time_after_intervention, arima(2,0,3))
I’ve got results from ARIMA:
Time - coef.
Intervention - coef.
Postslope - coef.
Cons –coef.
AR(L1-L2) 2 x coef.
Ma(L1-L3)-3 x coef.
Sigma - ( coeff.)
I know I should know this but I am unsure. Is it possible to interpret time coefficient similar to e.g linear regression – every month my dependent variable increases/decreases? I think it is not possible but lately I saw similar interpretation in a good journal article.
In stata manual it stays that the inference is possible only with covariance stationary process. Is inference also possible with trend stationary process?
Thank you in advance!
Elzbieta
I am a new statalist user.
I think I want to use arima model but I am unsure if it is ok for interrupted time series with deterministic trend and I have some difficulties in interpreting coefficients. I am using stata10 for windows.
I have monthly time series with deterministic trend. Time series is trend stationary. For my research, I am interested in changes in time trend after intervention. I know it can be done with Prais-Wintsen regression but it works only for AR(1) code: prais depvar time intervention time_after_intervention
I modelled ARIMA(2,0,3) process with time, intervention, timeafter intervention as independent variables (arima depvar time intervention time_after_intervention, arima(2,0,3))
I’ve got results from ARIMA:
Time - coef.
Intervention - coef.
Postslope - coef.
Cons –coef.
AR(L1-L2) 2 x coef.
Ma(L1-L3)-3 x coef.
Sigma - ( coeff.)
I know I should know this but I am unsure. Is it possible to interpret time coefficient similar to e.g linear regression – every month my dependent variable increases/decreases? I think it is not possible but lately I saw similar interpretation in a good journal article.
In stata manual it stays that the inference is possible only with covariance stationary process. Is inference also possible with trend stationary process?
Thank you in advance!
Elzbieta
