Stata codes for Unfolding bracketed responses

[Thilini Weeraratne]

Hello there, I am graduate student working with income and cost data. I have collected primary data about income and medical expenditure using 'bracketed survey questions' (commonly used in HRS dataset). I am seeking help to understand the procedure to unfold the responses. There are two steps to the process:

1. Based on the selected covariates find a fit for all the data (with exact responses and missing responses) using linear regression.
2. For each missing value in closed brackets, find the observation that is closest in predicted value, among the observations that report an exact amount within that bracket.

I am comfortable performing the step 1 in Stata, yet not step 2.

Appreciate help or advice on this procedure.

Thank You

[George Ford]

see if this helps:

ssc install hotdeck
help hotdeck

Nearest neighbor matching within brackets would work. (https://www.numberanalytics.com/blog...eck-imputation)

[Thilini Weeraratne]

Thank you, George, I will check on this.

[Stephen Jenkins]

Thilini Weeraratne Also check out interval regression (-help intreg-); it allows you to have either point or interval (bracketed) responses