SAMPLING - Food Resources Manual

All sites are monitored only when the number is very small. In most cases, conclusions extrapolated from visits to a sample of sites can be used to the validate the accuracy of the information provided by the site reports.

Statistical sampling attempts to strike a balance between the implausibility of completely examining the performance and transactions of all sites on one hand and the selection of a sample where the margin of error is within an acceptable range on the other hand. It is important that sample sites are selected from the master list of approved distribution sites and that sites are selected in such a way that every site has an equal chance of being selected.

Types of Statistical Sampling

The following general information and suggestions on selecting sample sizes, choosing a methodology and interpreting data can be augmented by further assistance from regional managers, Technical Assistance Group, CARE’ Internal Audit Department and other consultants.

Country offices must reach agreement with local donor representatives on methods of sampling, selecting sample sizes and interpreting data.

  1. Unrestricted Random Sampling
  2. This method assumes that each site has an equal chance of being part of the sample selected. Make a list of all project sites, perhaps by alphabetical order. Every project site is given a number. Once the total number of sites is known, decide how many sites are required for the. Use a table of random numbers to decide which site is selected first and the pattern for selecting sites thereafter. For instance, the table might tell you to start with Site #4 and select every 6thsite after that until a sample of 20 sites has been selected.

    Random sampling isn't always the most convenient method of choosing a sample. If there are many, many sites and the number of sites selected is small, the random method will almost always produce a sample across many different regions and terrain. It may not be physically possible, given the number of monitors, vehicles and fuel available to visit all the randomly selected sites in a prescribed time frame. For example, it is unrealistic to expect a monitor to witness distributions at two sites per day if s/he must travel hundreds of miles by motor bike or public transportation. Other types of sampling, such as stratified random or systematic may be more appropriate.

  3. Stratified Random Sampling
  4. This method of sampling is sometimes used if there are wide variations in site performance within a certain geographic location or type of distribution site (i. e., health centers or schools). All the sites are grouped into segments, each having some uniform, easily identifiable characteristics. Each segment is sampled separately using unrestricted random sampling methods. For instance, there might be a sample taken of all the school distribution sites and another sample taken of all the health centers. Within the segment, each site must have the same probability of being selected as any other site. At the end of the examination of each segment, the results from all segments are jointly evaluated.

  5. Systematic Sampling
  6. In systematic sampling, the selection plan is established by selecting a random start and setting a sampling interval that would result in choosing a previously specified sample size. For example, the third site on the list maybe the first site monitored and thereafter every tenth site will be included in the sample.

Interpreting Statistical Data

  1. Precision
  2. Project management must draw conclusions from the results of the sample.

    Because the sample may not show the true characteristics of the entire population of sites, a certain risk is involved in all samples. It is possible to quantify how much variation to expect as a result of errors under certain conditions, e.g., ± 2%.

  3. Margin of Error
  4. There are two types of error: sampling and non-sampling error. Non-sampling errors include listing errors and omission, response and measurement errors, errors of coding and data entry. Sampling error refers to errors that are attributable to the fact that the estimates are being made from the sample rather than testing the entire universe.

  5. Confidence Level
  6. This has to do with the percentage chance of drawing a correct conclusion from the sample. For example, a 95% confidence level means that there is a95% chance that the true value of whatever is being measured lies within the specified precision. In other words, there is a 5% chance that the true value for the population does not lie within the specified precision. Usually a larger sample size will result in a higher confidence level.

Selecting the Sample Size

There are a number of factors to consider when determining an adequate sample size.

First is a determination of the number of variables or factors which are expected to have a significant influence on systems management. Variables may include:

  • Available staff and support infrastructure (health posts vs. health centers)
  • Accessibility of site to supervision and supplies (urban vs. rural)
  • Type of institution (private vs. public, MCH vs. school feeding, community based or government)
  • Size of catchment area, i.e., geographical area and population served by the site
  • Amount of food and other resources being used in a project
  • Estimated amount of loss or current inventory in sites.

The actual number of sample sites to select will depend on what is being measured.

  1. Estimating Values
  2. If information on the actual amount of loss or inventory is required, sample sizes may be developed using the table below. Determination of this sample size is based on the general rule that the sample size must be high enough to allow for representation of each value to be estimated.

    Sampling Guidance

  3. Attributes Sampling
  4. Attributes sampling is a method used to estimate the proportion of specific attributes in a population. This proportion is called the occurrence rate and is the ratio of the attributes to the total number of the population. For example, country offices may be interested in knowing the percentage of centers complying with reporting requirements. Attributes samples vary only slightly with population size. For example, the sample size for a population of 500 is almost the same as the sample size for a population of 2000.

    This distinction is important because it may determine just how large a sample size must be drawn. If there are specific needs to look at, such as the actual size of a loss or the amount of damaged food shipped to centers, the total number of centers must be taken into account. On the other hand, for attributes sampling, a smaller sample size can provide managers with sufficient information to make informed decisions about how well distribution sites are complying with reporting requirements.

Cost Effectiveness

Early in the development of monitoring systems, country offices must consider the practical questions about the cost of monitoring activities including the time and travel of staff and staff support. Consideration must be given to:

  • Salaries and other personnel costs - program management, field staff, clerical and consultants
  • Travel
  • Office rent in the field
  • Vehicle purchases and maintenance
  • Supplies and equipment
  • Administration - printing, postage, telephone
  • Other costs - overhead.

Country offices must assure themselves that sample sizes are not larger than they can afford. If country offices do not have adequate personnel and resources to monitor the sample size required to insure a 95% confidence level, a lower confidence level, such as 80%, may have to be set. In these cases, country offices should inform regional managers and reach agreement with local donor representatives to assure that donor requirements on monitoring and sampling are satisfied.


All rights reserved © 2018 Wisdom IT Services India Pvt. Ltd DMCA.com Protection Status

Food Resources Manual Topics