exclude states without pre trend

add in competing method for overall fisher's for prevalent ae (prev>=0.2), try different prev

individual fisher with bonferroni correction (prev >=0.2)

try different sample size: 500, 1000 total to see if diff in power increases for binom 0.2 vs 0.25 (since it is more realistic)

delete this scenario

keep plot, try nsim 1000 nperm1000

findings: 1. when no noise, overallFisher and Unwt show highest power amd IndiFisher with Bonferroni shows lowest power 2. as # noises increase, power decreases

remove overlapping records and keep only the initial records

will also be used for pain management, Yongmei will send me the ICD code for identifying Pain patients, check if all Buprenorphine user in NonOUD cases are due to pain

will also be used for AUD, Silvia will send me the ICD code for identifying AUD patients, check if all Naltrexone user in NonOUD cases are due to AUD

might need to combine some insurance cat