* open data
use daily_mortality

* open log file
log using daily_mortality.log, replace

* generate a stata date (linear date integer)
gen date=mdy(month,day,year)

* define the data as a time series data set
tsset date


* get log of counts
gen lcounts=ln(counts)

* construct month and weekday dummy variables
xi i.month i.weekday


* construct variable for the 1st week of the month
gen firstweek=(day>=1&day<=7)
gen sept911=(month==9&day==11&year==2001)


* construct a monthly trend that equals
* 1 in jan of 1973, 2 in feb, etc
gen trend=month+12*(year-1973)

* run a regression, controll for
* time trend, month and weekday
* effects and first of the week
* the phrase _I* will include
* all variables with an _I prefix to
* the variable name
reg lcounts trend sept911 firstweek _I*

* get durban watson
estat dwatson

* output residuals
predict r, residual

* get 14 lags of residuals
gen r1=r[_n-1]
gen r2=r[_n-2]
gen r3=r[_n-3]
gen r4=r[_n-4]
gen r5=r[_n-5]
gen r6=r[_n-6]
gen r7=r[_n-7]
gen r8=r[_n-8]
gen r9=r[_n-9]
gen r10=r[_n-10]
gen r11=r[_n-11]
gen r12=r[_n-12]
gen r13=r[_n-13]
gen r14=r[_n-14]

* run regression of r on r1-r14
reg r r1-r14



log close