If you cluster your sample, first determine the number of clusters you will select.
The number of clusters (e.g., communities, villages, administrative units, groups)
should be decided based on the variability between clusters and the variability
within clusters. Aim to capture the greatest degree of variability within your sample.
If you anticipate that units within clusters (e.g., households within communities)
are relatively similar and that there is greater degree of variability between
clusters, you should opt for more clusters to capture the variation. In this case,
with a sample of 432 units, you could select 31 clusters (communities), and 14
units (households) within each cluster.
If you anticipate a greater degree of variability within clusters and the clusters
themselves are relatively similar to one another, select fewer clusters and more
units within each cluster. For example, you could select 14 clusters and 31 units
within each cluster. Note that both options result in a sample of 432 units.
If you have reliable population data for each cluster, follow guidelines
38
for probability
proportional to size cluster sampling. This is an important step to keep the data fully
representative of the survey population.
If you do not have reliable population data for each cluster, simply select the desired
number of clusters randomly.
Selecting sample units without a complete list of units
If you do not have a complete list of units, the teams will need to select the units
(e.g., households) to survey once they arrive in the clusters (e.g., villages).
Provide additional training on this selection method to ensure that all teams carry it
out in the same way to preserve the random nature of sample selection.
Team supervisors should be responsible for the selection process, but enumerators
may participate as well. To begin sampling households in a given village:
Find the center of the village. Work with the village leaders to define the
geographic boundaries and identify the center point. If you think there may be
poorer or marginalized households living at the edge of the village that you
would like to include in your survey, ask the leaders to include these areas
within the village boundaries for this exercise.
Stand at the center of the village. Spin a pen in the air and let it fall to the ground
to determine a random direction.
Walk to the edge of the village following the direction of the pen and count the
number of households that you find along this line.
Randomly select one of these households to start. For instance, if you counted
seven households along this line, randomly select one of these seven
households—this will be the first household to survey. One method to select the
first household randomly is to write the number of each household along this
line (one through seven in this example) on a separate piece of paper. Fold each
piece of paper so the corresponding household number cannot be seen and ask a
team member, or village leader, to randomly select one of the pieces of paper.
The number on the selected piece of paper will be the first household included in
the survey.
To select the next household to survey, look to your right when facing out of the
door of the first household. The first household in your line of vision will be the
next household to survey. Continue selecting households in this way until you
have reached your sample size.
If, following this method, you reach the edge of the village prior to completing
your sample, return to the center of the village and repeat the selection process.
Begin by spinning the pen to select another direction randomly and continue all
steps as indicated above.
Itis often more time-efficient for enumerators to begin collecting data while the
supervisor is still continuing the household selection process. If appropriate, the
supervisor can mark the selected households so that enumerators can head straight
to the next household for data collection. Common methods for marking
households include placing a colored piece of paper under a rock near the door or
making a chalk mark on the same rock. Do not mark houses if there is any chance
that this would be culturally inappropriate or would decrease the likelihood that
households would be willing to participate.
