## Presentation on theme: "Copyright 2010, The World Bank Group. All Rights Reserved. Agricultural Census Sampling Frames and Sampling Section A 1."— Presentation transcript:

Copyright 2010, The World Bank Group. All Rights Reserved. What is Sampling? Selecting a representative group of units from a population in order to make estimates about the population from the sample. 2

Copyright 2010, The World Bank Group. All Rights Reserved. What is a Sampling Frame? A list of all the units from which the sample will be selected. It can take various forms, depending upon the unit of analysis. 3

Copyright 2010, The World Bank Group. All Rights Reserved. Why Use Sampling? Saves time and money in comparison to complete enumeration Can concentrate on a particular sub-group of interest (e.g. livestock) Where nonsampling errors are large, sampling allows for better control 4

Copyright 2010, The World Bank Group. All Rights Reserved. When to Not Use Sampling If data are needed down to the lowest administrative level in a country Measuring small changes over time High survey costs related to sampling, field control, etc. 5

Copyright 2010, The World Bank Group. All Rights Reserved. Characteristics of a Good Sampling Plan Known chance of selection for each unit Measurable reliability Feasibility Economical and efficient 6

Copyright 2010, The World Bank Group. All Rights Reserved. What Size Sample do I Need? The most practical sample size that will accurately estimate a population parameter with a specified precision using the most efficient design. 7

Copyright 2010, The World Bank Group. All Rights Reserved. Systematic Sampling Basically as accurate as simple random sampling How to do it: –Determine your sampling rate –Use random number table to find a starting number equal or less than the sampling interval –From starting number take every nth unit 8

Copyright 2010, The World Bank Group. All Rights Reserved. Sampling Frame Consideration When a sampling frame contains units that are out of scope do not substitute: –Results in higher sampling rate –May introduce bias in the selection process 9

Copyright 2010, The World Bank Group. All Rights Reserved. Sample Stratification When characteristics of the population are known they can be used to improve sampling Similar sampling units are grouped together into strata A simple random sample selects at least one unit from each group or strata Must know size of strata and have reliable frame for selection of units from each strata 10

Copyright 2010, The World Bank Group. All Rights Reserved. Probability Proportional to Size Possible when a size measure is available that is correlated to the variable of interest Can improve accuracy by ensuring that large units which have the greatest impact on population estimates have the same probability of selection Used when sampling units vary in size 11

Copyright 2010, The World Bank Group. All Rights Reserved. Cluster Sampling Sampling units are grouped into clusters instead of a list and a sample of clusters is chosen rather than individual units. –Good for when no adequate frame exists of the units in the population –Or for saving field costs by cluster sampling on some geographic basis 12

Copyright 2010, The World Bank Group. All Rights Reserved. Single-Stage Cluster Sample A cluster of units is selected and all units in the selected cluster is sampled –Good for when there is no list of units or when sampling at the unit level is inefficient –Can also use area samples, where a geographic area is considered a cluster 13

Copyright 2010, The World Bank Group. All Rights Reserved. Multi-Stage Cluster Sample A subsample of units in the selected cluster is selected for the sample –A sample of clusters is selected –A list of all units in the selected clusters is constructed –A sample of units in the selected clusters is selected Reduces the listing required and 14

Copyright 2010, The World Bank Group. All Rights Reserved. Area Sampling Area sampling is a frequently used method of cluster sampling, used when: –Complete household lists are unavailable but reasonably good maps exist –Travel costs for covering randomly selected households is too great 15

Copyright 2010, The World Bank Group. All Rights Reserved. Single-Stage Area Sampling Obtain a detailed map and update with local authorities, noting changes on map Number blocks serially in a serpentine manner to avoid omissions Select a simple or systematic 1% sample of blocks Interview all households in the selected sample blocks 16

Copyright 2010, The World Bank Group. All Rights Reserved. Two-Stage Area Sampling Take one block in every 25 as a sample block instead of 1 in a 100 used in procedure A. Divide each sample block into 4 parts with approximately equal numbers of households Number each block segment from 1 to 4 Randomly select one of the four segments in each sample block and interview all households in the selected segment 17

Copyright 2010, The World Bank Group. All Rights Reserved. Area Sampling with Listing When there are no maps then: Select 1 in 25 blocks List and number all households in the sample blocks Select ¼ of all households in each sample block by simple random or systematic sample Interview all selected sample households 18

Copyright 2010, The World Bank Group. All Rights Reserved. Choice of Sample Design Sampling specialists must choose an efficient design to provide the desired precision balancing costs and standard error – When cost is not important, single-stage sampling provides more accurate results –When cost and administrative convenience are important a cluster sample is often used 19

Copyright 2010, The World Bank Group. All Rights Reserved. Section A Quiz 1.What are the characteristics of a good sampling plan? 2.What is area sampling and when is it used? 3.When should you not use sampling? 4.What is stratification? 20