Martijn van Hasselt, Christopher R. Bollinger, and Jeremy Bray, "A
Bayesian Approach to Account for Misclassification in Prevalence and
Trend Estimation," Journal of Applied Econometrics, Vol. 37, No. 2,
2022, pp. 351-367.

The data used in this paper are taken from the public-use files of the
2002-2014 National Survey on Drug Use and Health (NSDUH). Per the
terms of use, the data cannot be hosted in the journal archive. 
However, the combined file for the 2002-2014 time period can be
downloaded from the Substance Abuse & Mental Health Data Archive
(SAMHDA) at 
https://www.datafiles.samhsa.gov/dataset/national-survey-drug-use-and-health-2002-2014-nsduh-2002-2014-ds0001.


Main analysis variables: 

ANALWC1 = survey (population) weights when using data anually.

ANLYR = binary indicator for past-year, non-medical use of opioid pain
relievers. Non-medical use refers to use of medications that were
either not prescribed, or that were taken solely for the experience or
feeling they caused; "1" = used during the past year; "0" = did not
use during the past year.  


Sample selection variables: The variables used for sample selection and subgroup analysis are as follows.

AGE2 = categorical variable for age; >="7": selects all individuals 18
or older; "13" = 26-29 years; "14" = 30-34 years; "15" = 35-49 years.

IRSEX = gender indicator; "1" = male; "2" = female. 

NEWRACE2 = categorical variable for race; "1" = Non-Hispanic white.

COUTYP2 = categorical variable for metro status; "1" = large metro;
"2" = small metro; "3" = non-metro.


Please address any questions to
:
Martijn van Hasselt
Department of Economics
The University of North Carolina at Greensboro
P.O. Box 26170
Greensboro, NC 27402
mnvanhas [AT] uncg.edu