Consider this situation. 60 drug treatment sites. Dependent variable = % of attendees at each drug treatment site on Drug A. Propensity matched (for individual level and site level characteristics) control group created using data. Dependent variable collected every quarter for 2 years prior to intervention introduced at 60 sites and for 4 years after intervention. Want to know whether intervention has an impact on % taking Drug A. Is ARIMA the way to go here?
So your time series consists of only [(2+4=6) years x 4 quarters = 24 observations]? If it is the case, I don't think ARIMA can be applied to a such dataset.
As Ozgur mentioned, your time series consists of 24 observations which is not enough for time series based models. I suggest you to consider statistical analysis methods such as ANOVA to interpret your dataset.
Yes, I also think prediction methods, e.g. ARIMA are not suitable in this case; statistical methods, e.g. statistical tests for group comparison, would be more appropriate.