Introduction to Sampling : Censuses vs. Sample Surveys

1
Introduction to Sampling Censuses vs. Sample
Surveys

• Module 3
• Session 4

2
Session Objectives

• Distinguish between censuses and sample surveys
• Demonstrate the linkages between censuses and
  surveys
• Discuss the challenges of conducting censuses and
  large scale surveys in Uganda
• Distinguish between random and non random samples
• Identify the types and/or sources of errors in
  censuses and surveys
• Discuss how errors can be minimised in censuses
  and surveys

3
Reminder of Definitions

• Population totality of all units of interest
• Sample part/subset of the population
• Censuses inquiries that cover the whole
  population eg. Uganda Population and Housing
  Census, CIS, EMIS, HMIS, LOGICS, etc
• Sample surveys are inquiries that cover
  part/subset of the population eg. UDHS, UNHS,
  NSDS, etc
• Sampling Frame list of distinct and
  distinguishable units in the population of
  interest beginning step in almost all random
  sampling schemes, e.g. numbers written on
  households before the census night

4
Other Definitions

• Defacto census- covers all persons found within
  the borders of a particular territory/country at
  a particular point in time-census night
• Dejure census-tallies people according to their
  regular or legal residence

5
Sampling Frames

• Sources
• Administrative records-eg
• Hospital records
• Birth and Death Registers
• LC lists
• Voters register
• School registers
• etc
• Construct your own

6
Disadvantages of various sources of sampling
frames

• Administrative records may not be up to date
• Constructing your own may be too costly
  especially in large scale surveys

7
Role of censuses in Uganda

• Provide benchmark data for monitoring, planning
  and policy formulation eg we need data for
• UPE monitoring,
• poverty monitoring
• Election monitoring
• Resource allocation

8
Role of censuses in Uganda (cont.)

• Provide small area statistics - basic data
  disaggregated to the lowest administrative unit
  e.g we use census data to know the number of
  people in each village, sub county and district
  for planning purposes
• Show the actual status of the various indicators
• Health indicators-mortality, disease prevalence
• Fertility trends, population growth rate

9
Linkages between censuses and sample surveys

• Sample surveys can be used as a substitute for
  censuses
• Sample surveys can be used to supplement census
  data
• Sample surveys can be used to pretest census
  materials, procedures and methods
• Censuses are used as a basis for surveys
  conducted between censuses
• Sample surveys can be used to monitor census
  results

10
Challenges of Conducting Censuses and Large Scale
Sample Surveys

• Challenges of Surveys and Censuses Mubiru
  James.ppt

11
Types of Samples

• There two types of samples
• Random and Non random samples
• Random samples are those whose composition is not
  influenced by the sampler
• Non Random samples are those whose composition is
  influenced by the sampler

12
Advantages of Random Samples

• Objective and hence inferences based on them are
  reliable

13
Disadvantages of Random Samples

• Costly to select
• Need skilled manpower to get a random sample
• For some surveys, random sampling may not be the
  best because the sample may not provide the
  required data.

14
Advantages of Non Random Samples

• Easy and cheap to select since selection and
  substitution can be done at will
• Since they are done at will, the data needed can
  be easily obtained

15
Disadvantages of Non Random Samples

• Subjective and hence inferences based on them are
  biased
• Sampling errors can not be estimated

16
Types of Errors

• There are two types of errors, namely
• Sampling errors
• Non sampling errors

17
Sampling Errors/Biases

• Sampling errors are absent in censuses
• Their causes include
• Use of defective sampling frame
• Use of defective sampling procedures
• Use of an estimation method that does not
  correspond to the sampling design

18
Non Sampling Errors

• Non sampling errors occur both in censuses and
  sample surveys but are more pronounced in
  censuses

19
Sources of Non sampling Errors

• Defective sampling frames resulting into coverage
  errors
• Under coverage
• Over coverage
• Conceptual problems
• Physical environment
• Inadequacy of enumerators and supervisors

20
Sources continued

• Language problems translation
• Problems of measurement
• Response problems
• Non response problems
• Poor cartographic work
• Poorly designed questionnaires/instruments
• Poorly trained enumerators/supervisors
• Unqualified enumerators/supervisors

21
How Errors can be Minimised

• Supervision
• Training
• Use of the appropriate estimation method
• Publicity of the survey
• Testing the survey instruments

22
Sampling in the Research Process

• Problem
• Objectives
• Hypotheses
• Methodology
• Data Sources
• Target population
• Census or sample?
• If sample?
• What is the sampling design?