Hi Frank,

I would take a look into interrupted time series analysis (which can be done with R), probably by using generalized additive models to estimate a potential nonlinear time trend before the treatment or after the treatment. Generalized additive models can be don with a single time series (GAMs) as well as multiples (than it becomes GAMMs, generalized additive mixed models).

Kontopantelis, E., Doran, T., Springate, D. A., Buchan, I., & Reeves, D. (2015). Regression based quasi-experimental approach when randomisation is not an option: interrupted time series analysis. Bmj, 350, h2750. doi:10.1136/bmj.h2750

Sullivan, K. J., Shadish, W. R., & Steiner, P. M. (2015). An introduction to modeling longitudinal data with generalized additive models: Applications to single-case designs. Psychological Methods, 20(1), 26.

Zhao, A., Chen, R., Qi, Y., Chen, A., Chen, X., Liang, Z., . . . Li, W. (2016). Evaluating the impact of criminalizing drunk driving on road-traffic injuries in Guangzhou, China: a time-series study. Journal of epidemiology, 26(8), 433-439.

HTH

--Holger

Thank you! I will look more into those. I appreciate you linking these studies for me. Holger Steinmetz

I'd also suggest a type of segmented regression (doesn't really need to be an additive model, it can also be a linear model, what facilitated the interpretation fo the coefficients). The "break point" is fixed at t0 (start of treatment). You can use a mixel model or fit a model for each individual separately and then analyze the interesting fitted coefficients from these models.