Hi,

Would appreciate some help - I'm relatively new to Stata.

I've done an interrupted time series analysis based on a negative binomial model for rates, and am interested in the percentage change in the incident rate just prior to the intervention compared to that at the end of the study. I can get this using lincom and this is effectively an incident rate ratio as well, but the reviewer for the article relating to this work wants a 'percent change' value (with CI's) as they claim this helps the reader with interpretation. Is this somehow possible to do in Stata? - e.g. IRR = 0.51 (0.29, 0.88) -> can we somehow get this expressed as a 49% reduction with associated CI's? (I'm assuming it's not just a simple case of inverting the endpoints [< 1] - or subtracting 1 [> 1]).

The code I've used is:

Code:
nbreg unitssupplied seq, exposure(popn) dispersion(mean) irr
margins, at(seq=(26 60)) exp(predict(ir)*100000)
lincom _cons + 26*seq, irr // pre-intervention IR
lincom _cons + 60*seq, irr // post-intervention IR
lincom _cons + 60*seq - _cons - 26*seq, irr // pre-post IRR
where seq is my time variable (up to 60 months). The intervention was at 26 months. (For simplicity I haven't included the interaction term for the intervention.)

Thanks,

Paul