Hi there,
I am very new to ARIMA modelling. I am attempting to apply ARIMA modelling to examine trends in psychotropic medication over a ten year period (I have data at each quarter).
I have gone through the steps to identify potential models. Based on examination of graphs and formal Dickey Fuller and Phillips-Perrons tests, my data is non-stationary. These tests are then significant when applied on first-differencing of my outcome variable. I then examined the partial auto-correlation and auto-correlation functions to determine p and q, which suggest the following potential models:
ARIMA (5,1,0)
ARIMA (6,1,0)
ARIMA (7,1,0)
However, when I run each of these models to compare key parameters, non of the AR terms are significant.
Any advice on next steps or key texts would be greatly appreciated!
Finola
I am very new to ARIMA modelling. I am attempting to apply ARIMA modelling to examine trends in psychotropic medication over a ten year period (I have data at each quarter).
I have gone through the steps to identify potential models. Based on examination of graphs and formal Dickey Fuller and Phillips-Perrons tests, my data is non-stationary. These tests are then significant when applied on first-differencing of my outcome variable. I then examined the partial auto-correlation and auto-correlation functions to determine p and q, which suggest the following potential models:
ARIMA (5,1,0)
ARIMA (6,1,0)
ARIMA (7,1,0)
However, when I run each of these models to compare key parameters, non of the AR terms are significant.
Any advice on next steps or key texts would be greatly appreciated!
Finola
