Hello!
I am trying to find if there is an association between two interventions on infection rate over an 11 year time span. The on infection is available in monthly increments. There are two interventions 1-change in testing type and 2-change in reporting making hospitals responsible for the infection. I’m not sure if I should use Poisson or Time Series with transfer function?
Description of data: The incidence of infection initially decreased from 500 in 2006 to 400 in 2009 but increased to 475 in 2010 and continued to trend upward to 800 in 2013. In 2016 the incidence declined to 740. This increase in incidence from 2009 to 2013 coincided with the rise use of PCR testing in the VA system. The second intervention occurred in July 2012.
When I run the model using Time Series- PCR is significant, but the second intervention of reporting was not significant. I adjusted the data using a differencing function of 1 for seasonality. R2=0.999
When I run the data using Poisson – Link function of Log, both interventions are significant.
Does anyone have any incite to this difference?
