This folder contains the following:
(Updated 6/28/2021)

*******************************************************************************
1.  The results for the exposure model %IU-Age 5 with baseline specifications,
where individual outcomes are measured within age range 35-45. 
*******************************************************************************

  a) For each birth cohort, we use birth year and survey year to identify which
survey data observations were collected between age 35 - 45.

  b) For each birth cohort, we calculate the average outcome across these ages.

  c) By county x birth year x birth month, we total the number of
underlying observations from each county x birthyr x birthmo x survey year cell 
within age 35-45. This will be the new number of underlying observations per
cell, after we collapse to the county x birth year x birth month level. These
are used as weights in the regression.

  d) We collapse the data to the county x birth year x birth month cohort. We
also drop observations for which we are unable to observe them within the age
range 35 - 45 (so all cohorts born in 1950-54 are dropped).

  e)Finally, we run the baseline model with our new outcome. The only change to
the model is we no longer include survey year fixed effects.

The output is titled "z_composite_35_45.xls"

Note: the mean of the dependent variable (mdv) reported in the excel sheet,
 is no longer the mean from the control group (1950-1954). 
It is the mean of all cohorts in the analysis sample 
(since we don't observe the outcome for 1950-1954)

Example:
Take the cohort born in county A in birth year 1960 and birthmonth 2. Suppose
we collect survey data for them in 2004, 2005 and 2006.

First, we calculate that the cohort is 44 in 2004, 45 in 2005, and 46 in 2006.

To construct our new outcome within the age range, we take the average of the
z_composite_all collected in 2004 and 2005. 

Additionally, if we had 9 underlying observations in 2004, 1 in 2005, and 9 in
2006, we identify that 10 underlying observations were used to create the
average outcome across ages 35-45. This is the weight we use in the regression.

We collapse the data to the county x birth year x birth month level, and run
the baseline model with the new outcome.

*******************************************************************************
2.  The results for the exposure model %IU-Age 5 with baseline specifications,
plus added controls for age and age squared
*******************************************************************************
The output is titled "z_composite_age_controls.xls".

Can compare to original results: output/20180408/z_composite.xls, column E for
sheet "all", column C for sheets "wm", "wf", "nm", "nf"

*******************************************************************************
3.  The results for the exposure model %IU-Age 5 with baseline specifications,
where individual outcomes are measured at a specific age, for ages 30-35 
*******************************************************************************
Outcomes are constructed like in (1) but for a specific age.

The output is titled "z_composite_30.xls" when outcome is measured at age 30,
"z_composite_31.xls" when outcome is measured at age 31,
"z_composite_32.xls" when outcome is measured at age 32,
"z_composite_33.xls" when outcome is measured at age 33,
"z_composite_34.xls" when outcome is measured at age 34,
"z_composite_35.xls" when outcome is measured at age 35

