help pv                                                       also see:  pvtest
-------------------------------------------------------------------------------

Title

pv -- Estimation with plausible values

Syntax

pv [indepvars] [if] [in] [weight] , pv(varlist) [options]

options Description ------------------------------------------------------------------------- Main pv(varlist) plausible value estimates of the dependent variable cmd("command") estimation command to estimate the statistic, default is regress cmdops("options") options for the estimation command

VCE BRR (PISA) brr specifies Balanced Repeated Replication (BRR) be used for VCE rw(varlist) BRR replicate weights to be used fays(real) Fay's adjustment

VCE JRR (TIMSS PIRLS) jrr specifies Jackknifed Repeated Replication (JRR) be used for VCE jkzone(varname) Sampling zone variable jkrep(varname) Zone replicate weight variable

VCE Bootstrap bs specifies that Stata's bootstrap command be used for VCE bsops("options") options for Stata's bootstrap command

VCE Aggregation pisa specifies PISA's formula for aggregating VCE be used timss specifies TIMSS formula for aggregating VCE be used pirls specifies PIRLS formula for aggregating VCE be used -------------------------------------------------------------------------

fweights, iweights, aweights, or pweights are required; see weight. See pvtest for post estimation hypothesis testing. The default VCE utilizes the VCE reported by the command for each plausible value.

Description

pv estimates statistics when there are multiple estimates of the dependent variable (which are sometimes referred to as plausible values) by calculating the statistics for each estimate of the dependent variable and then calculating their average. This program can be used with any statistics estimation commands that accepts weights including regress, qreg, logit, mean, etc. It is specially (but not exclusively) designed to be used with the PISA, TIMSS and PIRLS student achievement datasets.

The VCE can be calculated in one of four ways. The default is to use an aggregation of the VCEs for each plausible value which is reported by the specified stata command. Alternatively, Balanced Repeated Replication (used with PISA), Jackknifed Repeated Replication (used with TIMSS and PIRLS), or bootstrapping can be used.

See the Formulas section below.

Options

+------+ ----+ Main +-------------------------------------------------------------

pv(varlist) the variable list of plausible values or estimates of the dependent variable.

cmd("command") specifies the name of estimation command to be used. Estimation commands must allow weights, and they must output their estimates to e(b).

cmdops("options") options to be specified for the estimation command.

+---------+ ----+ VCE BRR +----------------------------------------------------------

brr specifies that Balanced Repeated Replication be used for estimating the sampling variance.

rw(varlist) specifies the variable list of replicate weights to be used. In PISA 2000, 2003, and 2006 these variables are w_fstr1 - w_fstr80.

fays(real) is the Fay's adjustment. In PISA 2000, 2003, and 2006 this is 0.5.

+---------+ ----+ VCE JRR +----------------------------------------------------------

jrr specifies that Jackknifed Repeated Replication be used for estimating the sampling variance.

jkzone(varname) is a categorical variable that specifies the sampling zone; for TIMSS and PIRLS this variable is jkzone.

jkrep(varname) is a binary variable that specifies the observation's weighting within its zone; for TIMSS and PIRLS, this variable is jkrep.

+---------------+ ----+ VCE Bootstrap +----------------------------------------------------

bs specifies that bootstrapping should be used to calculate the sampling variance. The expression list is assumed to be e(b); as a result, the user can only control the options.

bsops("options") specifies any bootstrapping options; see bootstrap.

+-----------------+ ----+ VCE Aggregation +--------------------------------------------------

pisa tells the program to aggregate the VCE according to the formula on page 79 of OECD (2005a); this is the default.

timss tells the program to aggregate the VCE according to the formula on 2-52 of IEA (2005).

pirls tells the program to aggregate the VCE according to the formula on 2-52 of IEA (2005); see the description of formulas below.

Examples

Estimating statistics with PISA

. pv stratio propqual [aw=w_fstuwt], pv(pv*math) cmd("reg") brr rw(w_fstr*) fays(0.5)

. pvtest stratio=propqual

. pv stratio [aw=w_fstuwt], pv(pv*math) cmd("qreg") brr rw(w_fstr*) fays(0.5)

. pv stratio [aw=w_fstuwt], pv(pv*math) cmd("mean") brr rw(w_fstr*) fays(0.5)

. pv stratio [aw=w_fstuwt], pv(pv*math) bs bsops("reps(80) str(stratum) cl(schoolid)")

Estimating statistics with TIMSS and PIRLS

. pv [aw=tchwgt], cmd("mean") pv(asrrea*) jkzone(jkzone) jkrep(jkrep) jrr timss

. pv asbgsex [aw=tchwgt], pv(asrrea*) jkzone(jkzone) jkrep(jkrep) jrr timss

Estimating statistics with BRR VCE with a regular dependent variable

. pv stratio [aw=w_fstuwt], pv(propqual) cmd("reg") brr rw(w_fstr*) fays(0.5)

Formulas

To calculate statistics that are functions of plausible value estimates of a variable, the statistic is calculated for each plausible value and then averaged. See OECD (2005a), page 79 for the formula used in this program.

To calculate the standard error of statistics that are functions of plausible value estimates of a variable average sampling variance of the statistic for each plausible value combined with the variation among the estimated statistics for each plausible value.

By specifying the pisa option, the formula used to calculate the standard error can be found on page 79 of OECD (2005a). By specifying the timss or pirls option, the formula can be found on page 2-52 of IEA (2005).

The formula for calculating the VCE for statistics using BRR is taken from page 51 of OECD (2005a) while the formula used for calculating the VCE for JRR is from page 2-50 in IEA (2005).

The formula for calculating the parameter for the t-distribution of the ratio of a statistic and its standard error is on page 131 of OECD (2005b).

References

Lauzon D. (2004), "Variance estimation with plausible value achievement data: Two STATA programs for use with YITS/PISA data", Information and Technical Bulletin, Statistics Canada, Ottawa

OECD (2005a), PISA 2003 Data Analysis Manual, OECD, Paris

OECD (2005b), PISA 2003 Technical Report, OECD, Paris

IEA (2005), TIMSS 2003 User Guide for the International Database, IEA, Chestnut Hill, MA

Author

Kevin Macdonald is a consultant with the Education Team in the World Bank's Human Development Network and has a master's degree in economics from the University of British Columbia in Vancouver, Canada. Please send questions, comments or complaints to kadmacdonald@gmail.com.

The command syntax and display for this program was inspired by Lauzon (2004).

Also see

pvtest

Version

Last updated 2008-09-05