{smcl}
{* 29apr2011}{...}
{cmd:help rrreg}{right:also see: {helpb regress}}
{hline}
{title:Title}
{p 4 4 2}{hi:rrreg} {hline 2} Linear probability model for randomized response data
{title:Syntax}
{p 8 15 2}
{opt rrreg} {depvar} [{indepvars}] {ifin} {weight} [{cmd:,} {it:options}]
{synoptset 26 tabbed}{...}
{synopthdr}
{synoptline}
{synopt :{opt pw:arner(#|varname)}}probability of the non-negated question
in Warner's model; default is {cmd:pwarner(1)}{p_end}
{synopt :{opt py:es(#|varname)}}probability of a surrogate "yes"; default is {cmd:pyes(0)}{p_end}
{synopt :{opt pn:o(#|varname)}}probability of a surrogate "no"; default is {cmd:pno(0)}{p_end}
{synopt :{it:{help regress:regress_options}}}options as described in help {helpb regress}{p_end}
{synoptline}
{p2colreset}{...}
{p 4 6 2}
Prefix commands allowed as described in help {helpb regress}.
{p_end}
{p 4 6 2}
{opt aweight}s, {opt fweight}s, {opt iweight}s, and {opt pweight}s are allowed; see {help weight}.
{p_end}
{title:Description}
{pstd} {cmd:rrreg} fits a linear probability model to data collected using the
randomized response technique (RRT). {it:depvar}=0 indicates a negative
outcome (a "no" answer); {it:depvar}!=0 & {it:depvar}<. (typically
{it:depvar}=1) indicates a positive outcome (a "yes" answer).
{pstd}{cmd:rrreg} is suited for the analysis of data from Warner's RRT scheme,
the forced-response RRT, or the unrelated-question RRT with a known
distribution for the non-sensitive question (see, e.g., Fox and Tracy 1986).
{title:Options}
{phang} {opt pwarner(#|varname)} specifies the probability of the non-negated
question in Warner's RRT scheme. {it:#} must be in [0,1]
and may not be 0.5. The default is {cmd:pwarner(1)}. Individually varying
probabilities may be specified by {opt pwarner(varname)}, where {it:varname}
holds the individual probabilities.
{phang} {opt pyes(#|varname)} specifies the probability of a surrogate "yes"
answer. {it:#} must be in [0,1]. The default is {cmd:pyes(0)}. Individually
varying probabilities may be specified by {opt pyes(varname)}, where
{it:varname} holds the individual probabilities.
{phang} {opt pno(#|varname)} specifies the probability of a surrogate "no"
answer. {it:#} must be in [0,1]. The default is {cmd:pno(0)}. Individually
varying probabilities may be specified by {opt pno(varname)}, where
{it:varname} holds the individual probabilities.
{phang}
{it:regress_options} are options as described in help {helpb regress}.
{title:Examples}
{com}. rrreg infidelity male age, pyes(0.5){txt}
{title:Methods and formulas}
{pstd}The randomized response regression model is estimated by fitting
a linear regression to a transformed dependent variable
{it:depvar} - (1 - {it:pyes} - {it:pno})*(1 - {it:pwarner}) - {it:pyes}
{it:y} = ----------------------------------------------
(2*{it:pwarner} - 1) * (1 - {it:pyes} - {it:pno})
{pstd}where {it:pwarner} is the probability of the negated question in
Warner's scheme and {it:pyes} and {it:pno} are the probabilities of a
surrogate "yes" and a surrogate "no".
{pstd}{it:pyes} and {it:pno} are unconditional (overall) probabilities. For
example, in an unrelated-question RRT where the probability to be directed
to the non-sensitive question is 0.4 (i.e. the probability to answer the
sensitive question is 60%) and the probability to answer "yes" to the
non-sensitive question is known to be, say, 0.75, the overall probability
of a surrogate "yes" is {it:pyes} = 0.4*0.75 = 0.3. Likewise, the overall
probability of a surrogate "no" is {it:pno} = 0.4*(1-0.75) = 0.1.
{pstd}{it:pwarner}, however, is conditional, i.e. it only applies to respondents
that are not instructed to give a surrogate "yes" or "no". That is,
overall (1-{it:pyes}-{it:pno})*{it:pwarner} percent of respondents answer the
original sensitive question, (1-{it:pyes}-{it:pno})*(1-{it:pwarner}) percent answer
the negated question.
{pstd}In the unrelated question design, if the distribution of the answers to
the non-sensitive question is unknown, an estimate has to be used to determine
the probabilities. {cmd:rrreg}, however, assumes {it:pyes} and {it:pno} to be
non-stochastic. To account for the additional variance induced by stochastic
{it:pyes} and {it:pno} you can, for example, apply {helpb bootstrap} to the
combined procedure of estimating {it:pyes} and {it:pno} and fitting
{cmd:rrreg}.
{title:References}
{phang}Fox, James Alan, and Paul E. Tracy. 1986. Randomized response: A method for sensitive
surveys. London: Sage.
{title:Author}
{p 4 4 2} Ben Jann, Institute of Sociology, University of Bern, jann@soz.unibe.ch
{title:Also see}
{psee}
Online:
{helpb regress}, {helpb bootstrap};
{helpb rrlogit} (if installed)