In our blog on Quantitative and Qualitative Data Collection Methods in Monitoring and Evaluation, we discussed that data sampling in qualitative methods is deliberately done for the most interesting cases while in quantitative methods, it is done for large, random samples. But what is sampling?
To produce a general picture of progress or impact, sampling refers to the selection of some, but not all, members of a target group as responders. There are two different sorts of sampling techniques.
- Random sampling is the process of selecting respondents at random, regardless of their location, gender, or any other important factor.
- Purposeful sampling is the process of selecting individuals or groups based on key individual or household characteristics (for example, whether they are male or female, engage in certain livelihood activities, or have critical vulnerability traits).
When should you not sample during an emergency response?
You should not sample when you need to count and document the actual number of products or services delivered (activity or output-level indicators). You must demonstrate real use of resources received and timely delivery of goods or services at the activity and output levels in order to satisfy donors. This necessitates precise and full data, which necessitates counting each supplied input or individual served. Begin counting as soon as the activity begins and continue until you reach the desired result and the activity is completed. It’s important to note that when an indication refers to a number, it’s a sign to count rather than sample. This is frequently the case for indicators at the activity and production levels.
When should you sample during an emergency response?
- As part of project monitoring, to assess the relevance and effectiveness of assistance.
Sample to check that intermediate results–level and strategic objective–level changes are starting to occur, to detect any problems that arise, and to inform ongoing management decisions if input from a small number of respondents is good enough to confirm that the community is satisfied with the assistance provided.
In an emergency, you can utilize a purposive sample to pick a limited number of community members to learn more about a certain subgroup’s perspective or why a change is occurring or not occurring. You might even talk to different subgroups on purpose to triangulate responses and gain a better understanding of the problem. You just need a modest number of respondents from each segment when using deliberate sampling. Depending on the method employed and the circumstances, the number might be as low as two or three groups or as many as eight or ten persons or homes.
A light random sample can be used to track satisfaction or behavior change at the output and intermediate results levels. A light random sample does not necessitate a set number of respondents; the number sampled is determined by how readily it can be included in the field staff’s work schedule, considering their many other obligations. Staff should select respondents as randomly as possible in order to reduce bias. This sample will not provide statistically accurate data, but it may be sufficient for identifying and addressing problems quickly.
- During evaluations, to determine and document impact against IR and SO-level metrics.
To illustrate project impact among the overall target population, use a representative random sample. When the number of respondents is calculated using an internationally known sampling equation, a random sample of respondents is representative of the target population and statistically significant. However, surveying a statistically meaningful random sample takes time, so it’s usually only done at the end of the project to quantify the impact and report to donors. When an indication refers to a percentage, it’s a sign that you should utilize a random sample. This is frequently the case for indications at the IR and SO levels.
Examples of sampling during an emergency response.
- Output level Indicator: The total number of hygiene kits supplied in a given period of time.
Count the number of kits you’ve distributed and keep track of it in your distribution records. A full count (no sampling) should commence as soon as the kit distribution begins and end only when the distribution is completed.
No evaluation is needed.
- IR level indicator: Percentage of the target population who use and store water correctly.
Check using a representative sample of women (who are generally responsible for water handling in the household). Select two or three of the worst-affected communities per month and, in each:
- Hold a focus group discussion with women.
- Observe water-handling behaviors in ten houses within target villages, chosen as randomly as feasible.
Use a random sampling of all target women at the midpoint or at the end to provide an accurate picture of the amount of acceptable use.
- IR level indicator: Sphere standards for risk reduction, comfort, and durability as met by a certain percentage of targeted households reconstructing shelters.
Use a quality-control checklist for shelter construction. Select the most vulnerable homes with care (those least likely to meet the indicator). Based on human resources and accessibility, determine what level of coverage and frequency is realistic.
Conduct a random sample survey of all targeted households at the conclusion of the initiative.
- IR level indicator: The percentage of people who were completely happy with the NFI kits they received.
Exit interviews with a small random sample of targeted community members, (e.g., 1 in 10) at a few distribution sites. Determine a sufficient number of sites based on available human resources. Consider including a household-level observation of NFI use in around 10 households in two to three of the worst-affected villages.
Note: You might wish to add a feedback mechanism to this light monitoring to spot problems that were missed by the short sample.
Conduct a satisfaction survey utilizing a random sample at the midpoint or at the end of the semester to provide a representative picture of the overall level of satisfaction with the response.
To learn more about MEAL data collection, register for a Monitoring, Evaluation, Accountability & Learning (MEAL) programme today.