Confidence interval that includes the value of zero
Opened this issue · 1 comments
Hello everyone, again I want to ask for your help.
I am estimating with multysynth but my confidence interval on the graph includes the value of 0, is this enough proof that the intervention has no significant effect on treated units despite the synthetic control fitting very well and despite that there is a noticeable trend of impact after treatment?
In the package examples I saw that for augsynth there were ways to improve the confidence intervals, changing parameters in the main function or in the summary. Are there also options to improve confidence intervals in multysynth?
Thank you very much
I can't answer specifically about the package. However, I would consider the width of your confidence intervals before suggesting that your intervention has had no effect. If your confidence intervals include large positive and negative effects, I would be more inclined to say that is indicative of noisy data and statistical uncertainty rather than the absence of any effect of the intervention.