I intend to do an interrupted time series to measure the adherence to a certain guideline. In time series, we need to decide the time till the effect of the guideline is expected or implemented. It was reported that this is done through looking at the literature. Does that mean looking at previous studies for the trend of the outcome in relation,and see if it is gradual change or abrupt change? If the implementation is expected to be gradual, does this means months or more?
Thank you Matteo for your answer, but it seems my question was not very clear. I intend to do a segmented autoregression analysis of the effect of of a certain clinical practice guideline on the adherence of family physicians, I'm asking about the time lag that should be considered in such analysis.
One of the easiest method to determine the number of lagged variables to be included in the model is the T-test implementation.You should consider the sequence of models with increasing number of lagged variables (start with just t, after t and (t-1), after that t, (t-1), (t-2), and so on) and when the T-test gives you some insignificant parameters corresponding to certain lagged variables you should stop and turned back to previous model with all significant parameters. But, of course, in the real cases the number of lagged variables could be strongly defined by the problems restrictions, or should be based on some theoretical background.
- Badges
- Science topic
- Similar topics
- Medicine
- Public Health