Programme Performance Assessment in Results-based Management

Session 3 - Collecting Data by Method: Surveys

Sampling is a technique of selecting a smaller number of cases from a larger group in such a way that they will represent the larger group within a known range of error. There are sampling techniques with various degrees of complexity. The simplest technique, randomly drawing a number of cases from the whole, like drawing numbers out of a hat, is the one on which all calculations of probable error are based.

Another simple method that is particularly applicable is to take every xth document using a random starting point until a pre-determined number of cases are selected. Suppose you want to examine only 50 scientific reports out of 500 that you have identified. This means you need to select one in ten. You can draw a random number between one and ten (from a hat, using a random number generator or pointing with your eyes closed at a table of random numbers) and then take every 10th document thereafter.

To get the largest reduction in error, stratified sampling is often used. The sample is stratified by classifying its composition according to factors that are known and are considered important in explaining variance, such as geographical region, type of State, type of institution and the like. Third, draw the sample, using random methods. Usually, the number drawn from each stratum is proportionate to the size of the stratum in the whole population.

For example, if potential questionnaire recipients from a training program classified as being in the Agriculture Ministry make up 25 percent of the total, then 25 percent of the sample should be drawn from that stratum.

For example, if an important consideration is the year that that a person was trained (due to changes in curriculum or an assumption that conversion of training into practice may take more time), and 10 percent of the trainees were trained in the first year of the project, then 10 percent of the sample should be drawn from this stratum.