Use random sampling when collecting quantitative data
Random sampling, also known as probability sampling, is statistically representative
of a survey population.
In other words, an appropriate random sample allows us to
survey a number of households and to generalize our findings to describe the larger
target population, including those households that have not been surveyed. By
definition, in a random sample every unit (within the target population) has an
equal chance of being selected. By ensuring this equal chance of selection, you are
able to generalize the results of the survey to the larger population.
Avoid any bias that is introduced into the sampling process (meaning that there is
evena slight difference in chances of selection), as it will question the ability of the
results to represent the larger target population.
Determine your sampling unit. Your sampling unit is your unit of comparison.
What or who will your data represent? Common sampling units include households,
women of reproductive age and children under age 5.
o Think ahead to your results. Refer to the indicators in your M&E plan to identify the
sampling unit you need to represent. If your sampling unit is a household, you cannot
state, ―16 percent of women of reproductive age reported receiving antenatal care
during their last pregnancy.‖ Conversely, if your sampling unit is women of
reproductive age, you cannot state, ―82 percent of households reported that their main
drinking water source is within 5 km of their home.‖
Identify the population that the sample will represent. Surveys need to limit their
scope to a certain area and often to specific households, or individuals, within that
area. You can determine the population by a combination of geographic boundaries
and household demographic characteristics or other inclusion and exclusion criteria.
Provide clear inclusion or exclusion criteria to define your sample population.
Ensure that all sampling units that fit these criteria will be represented by your data.
Consult your analysis plan when determining your sample population. Will your
data represent a district, a country or a community? What types of households or
individuals will your sample represent within this area? For example, are you
interested in representing the situation of project participants only or that of all
residents (participants and nonparticipants) in a certain geographic area?
Select the number of sampling units required by your sample size (refer to
Standard 2 below). There are two options for selecting units (such as households),
whether within clusters or within the overall sample population.
If you cluster your sample (discussed in Standard 2 below), refer to Annex A for
guidance on selecting clusters prior to selecting individual units.
If you have a complete list of all sampling units, follow the instructions in Annex B
for systematic random sampling. The list of sampling units must be up todate and
include all sampling units within your sample population. For example, your
sampling frame could be a list of all households in a given district or it could be all
women of reproductive age within designated districts.
Make sure your list includes all potentially marginalized households, individuals or
units that may be excluded from certain government or community lists.