FRAMES FOR AGRICULTURAL CENSUS AND SURVEYS

1
FRAMES FOR AGRICULTURAL CENSUS AND SURVEYS
2
INTRODUCTION

• The Samoan households survey framework
  incorporates the following sampling schemes
• ? Stratified sampling.
• ? Two Stage sampling
• ? Probability Proportional to Size sampling
• ? Systematic sampling

3
Stratified Sampling

• The Stratum adopted for the survey are the 4
  Statistical Regions which make up Samoa
• ? Apia Urban Area
• ? North West Upolu
• ? Rest of Upolu
• ? Savaii

4
Two Stage Sampling

• The two stages of the survey are as follows
• Stage 1
• Selection of enumeration areas
• Stage 2
• Selection of households within the selected
  enumeration areas

5
Probability Proportional to Size Sampling

• For the first stage the EAs are selected based
  on their size
• The size measure is the number of h/hs in EA
  according to the frame
• An EA with 100h/hs will therefore have as much
  chance of selection as an EA with 50h/hs

6
Systematic Sampling

• The second stage of selection, the households are
  selected using Systematic skip

7
The Cluster Size

• An early question to address in setting up the
  Framework is to determining how many households
  do we need to select per selected EA
• This is commonly referred to as the cluster size
• This is generally determined at the stratum level

8
The Cluster Size (cont)

• For Samoa, no technical work was undertaken to
  determine the most optimal cluster size.
• Instead we plagiarized results from other
  countries
• This resulted in a cluster size for each region
  of
• AUA 5
• NWU 10
• ROU 10
• SAVAII - 10

9
How many H/hs do we Select ?

• Best illustrated via an example (AUA) only
• ? Say we desire a 10 sample
• ? Total H/hs in AUA 5182 (census06)
• ? This mean for AUA, we need a Sample size of
  (10/100)5182518 h/hs

10
Selecting the Villages and EAs

• Step1
• Order villages by their geographical location
  within each Region
• Step 2
• Attach the total EAs for each village as well as
  the the total households for each EAs
• Step 3
• Convert the count of h/hs to the number of
  clusters by dividing the total h/h by cluster
  size for each Region. Eg (AUA5)
• Step 4
• Calculate the Cumulative number of clusters and
  add to the list for each Region

11
Selecting the Villages and EAs (cont)

• Step 5
• Determine the skip to be run through the list
  for each Region
• (total clusters in each region)/(selections
  for the region)
• Step 6
• Select a Random Start between the value of the
  skip and 0
• Step 7
• Produce the remaining selections numbers by
  continuously adding the Skip to the random start
  until you reach the total number of clusters in
  the stratum
• Step 8
• To determine the selected villages, locate the
  village with the lowest cumulative clusters
  value which is higher than the selection number
• Note At the same time the EAs is also selected
• (NB there may be 2 or more selections from
  village)

12
Selecting the Households

• On the household list of the selected village and
  EAs assign a cluster number to each h/h according
  to the Cluster Size
• Continue until the number of the clusters has
  been reach and then start at 1 again and repeat
  the process

13
Selecting the Cluster

• To determined which h/h or cluster to be
  selected, Subtract the selection number from the
  previous Cumulative number.
• Household list

14
For Excel file
15
1999 Census of Agriculture Framework
16
FAAFETAI LAVA !!!