Wash-in period ITSA

[Silje Olsen]

Dear Statlisters.
I am using the itsa command to perform an interrupted time series analysis with two interventions. I wish to incorporate a wash-in (out) period after my interventions. Is there any way to incorporate this into the itsa command? If not, how do I go about the issue?

Thank you!

[Jared Greathouse]

I don't understand. Could you talk about what this is and why you'd want a wash out period?


And please, do provide your data using dataex

[Silje Olsen]

Hi Jared. Thank you for your swift reply!

I have a dataset of time periods and corresponding prescription rates for antibiotics. My intervention is a drug shortage of a preventive drug. I want to investigate the effect of shortage of preventive drug on prescription rates for antibiotics. I will perform two single itsa, one for each line of antibiotics (A and B in dataset below) My assumption is that the intervention will have an immediate effect for some (those who were about to renew prescriptions) and increase in strength as more participants will experience shortage. As a sensitivity analysis, I want to add a wash out period after the intervention. I want the wash out period to be one time period after the intervention.

input float(time_period A B)
1 45.78904 8.830744
2 46.38866 8.122104
3 44.0447 9.648406
4 43.0635 10.68411
5 53.36604 10.575089

Do you have any input on how to code this?

[Jared Greathouse]

I understood that, my question is more like "What is a wash out period and why do you wanna model it?" I have an idea, but I wanna make sure I'm thinking of the right concept before I put it into code, so can you talk about what the wash out period is meant to represent first?

Also, is this your full dataset? it seems quite short. If not, please give the full/representative dataset for both units, and I'll get to seeing how I would think about it.

[Silje Olsen]

Ah, apologies, I misunderstood your question! By putting in a wash out period you want to do the its analysis, but without the data for a certain time period after the intervention - the period before the "intervention is assumed to take full effect". For this analysis do we assume that the intervention will have immediate effect for some (but not all), and then increase in effect. But I will like to add in a wash out period for one of my sensitivity analysis to check my results after the intervention from the time period where we as an alternate hypothesis assume "most will have the effect". My full dataset is 56 time periods long. I have two interventions, one in time period 27 and one in time period 34. I want the wash out period to be one time period after each intervention. I am only doing sensitivity analysis with wash out period on variable series A (you can ignore B for this code). Do you need me to post the full dataset or is that sufficient information? It just goes on like the example dataset until time period 56.( Not sure I am comfortable posting my full dataset online!)