Docstoc

PowerPoint Presentation - UNECE

Document Sample
PowerPoint Presentation - UNECE Powered By Docstoc
					Evaluating imputation of sex
  and age for substitutes
 in substitute households
                               Michael Ryan
     2008 UNECE Work Session on Statistical Data Editing
Overview
• The New Zealand Census of Population and
  Dwellings
  – Background
  – Dealing with non-response
     • Imputation procedures
• Estimation of census coverage
  – Post-enumeration Survey (PES)
The New Zealand Census
• Background
  – Population size: 4.2 million
  – Number of households: 1.5 million (in private
   dwellings)
  – Census
    • Frequency: every 5 years
    • Organisation
       – Types of forms
       – Data collection process
The New Zealand Census
• Dealing with non-response
  – Within census processes – unit imputation
    • creation of substitute dwelling forms
    • creation of substitute individual forms
    • imputation of sex and age
  – Outside of census – census coverage
    survey (Post-enumeration Survey – PES)
    • sample survey to estimate census undercount
      and overcount
Census imputation
• Creation of substitute households
  – Characterisation of a missing dwelling
     •   Listed in the dwelling frame
     •   Classed as a private dwelling
     •   Classed as occupied on census night
     •   Have no forms received for it
  – Errors caused by
     • Collectors
     • Processing
Census imputation
• Creation of substitute individuals within
  substitute households
  – Donor imputation
     • Random selection
     • From within the immediate physical neighbourhood
Census imputation
• Imputation of sex and age for substitute
  individuals in substitute households
  – 2001 Census
     • Stochastic imputation from the estimated resident
       population
     • Sex imputed first (at 51% female to 49% male)
     • Age imputed, conditional on imputed sex
Census imputation
• Imputation of sex and age for substitute
  individuals in substitute households (cont.)
  – 2006 Census
     • Imputation using the dwelling form
Census imputation
• Imputation of sex and age for substitute
  individuals in substitute households (cont.)
  – 2006 Census (cont.)
     • Donor imputation for age (from donor
       household)
     • Stochastic imputation for sex, independently of
       age
Post-enumeration Survey (PES)
• Objective
  – Measurement of overcount and undercount
    in census
• Additional benefit
  – Evaluation of imputation in census
Post-enumeration Survey (PES)
• Characteristics
  – Carried out as independently as possible
    from the census
  – Starts 2 weeks after the census
  – Runs for 2 weeks
  – Sample size – 11,000 households
  – Asks for addresses where people were on
    census night
Post-enumeration Survey (PES)
• What it can do
  – can identify people for whom substitute
    forms were created
  – can estimate
    • number of substitute households that should
      have been created
    • number of substitute individuals that should
      have been created, and their distribution by sex
      and age
Comparison of the number of substitutes
• Comparison between the number of
  substitutes in Census and PES
  – Test statistics used
             (C – P) / SE
    where
     C = Census count
     P = Post-enumeration Survey estimate
     SE = Sample errors (correspond to 95% intervals)
  – SE estimation
     • Jacknife methodology using GREG estimates
     or
     • Based on calculated design effect
Comparison of the number of substitutes
Number of substitute households and substitute individuals
 Comparison of the number of substitutes
Age and sex of substitute individuals in substitute households
Recommendations for 2011 Census
• That the rigorous checking of the Dwelling
  Frame against the collectors’ fieldbook that
  was used in 2006 be continued
• That further work be done to improve the
  classification of dwellings as occupied /
  unoccupied
• That the current donor household
  methodology continue to be used to impute
  the number of census night occupants
Recommendations for 2011 Census
• That sex for full-substitutes be imputed jointly
  with age
• That the possible use of PES 2011 estimates
  of full-substitutes to calibrate substitutes in
  substitute households be investigated in 2011
  Census
• That the current donor household
  methodology continues to be used to impute
  age for substitutes in substitute households
Contact
• Michael Ryan
  – Michael.Ryan@stats.govt.nz

				
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
views:2
posted:3/27/2013
language:English
pages:18