*unadjusted IR*
ir Count EthnicityWhite0 PersonYears
mepoisson Count EthnicityWhite0, exposure(PersonYears) vce(cluster LSOA) irr
*adjust for age gender, then deprivation*
mepoisson Count i.EthnicityWhite0 AgeCategory i.Genderf0, exposure(PersonYears) || LSOA: , vce(cluster LSOA) irr
mepoisson Count i.EthnicityWhite0 AgeCategory i.Genderf0 c.IMD, exposure(PersonYears) || LSOA: , vce(cluster LSOA) irr
*complete model with interaction term*
mepoisson Count i.EthnicityWhite0##c.EthnicDensity_Pakistani AgeCategory i.Genderf0 c.IMD, exposure(PersonYears) || LSOA: , vce(cluster LSOA) irr
*Complete model with area level random effects equation*
mepoisson Count i.EthnicityWhite0##c.EthnicDensity_Pakistani AgeCategory i.Genderf0 c.IMD, exposure(PersonYears) || LSOA: , irr
*examine interaction by splitting into high or low density*
*generate high low density and quartiles*
mean EthnicDensity_Pakistani
tab EthnicDensity_Pakistani
gen high=1 if EthnicDensity_Pakistani>12.79
replace high=0 if EthnicDensity_Pakistani<12.79
tab high
mean EthnicDensity_Pakistani, over(high)
by high, sort : mepoisson Count i.EthnicityWhite0 AgeCategory i.Genderf0 c.IMD, exposure(PersonYears) vce(cluster LSOA) irr
*compare by Pakistani quartiles*
drop if EthnicityWhite0==0
xtile density4 = EthnicDensity_Pakistani, nq(4)
mean EthnicDensity_Pakistani, over(density4)
mepoisson Count b4.density4 AgeCategory i.Genderf0 c.IMD, exposure(PersonYears) vce(cluster LSOA) irr
*compare within White Quartiles*
*reload dataset and use EthnicDensity_White equivalent in previous step*