Simple Random Sampling Explained With a Real Research Example
Simple random sampling is a probability-sampling method in which eligible units are selected from a defined sampling frame using a genuinely random procedure. Researchers do not choose participants because they are convenient, cooperative or easy to reach. To use simple random sampling well, you need a suitable sampling frame, a reproducible random-selection process and a plan for handling non-response without quietly replacing selected participants.
“Participants were randomly selected based on availability.”
No.
That sentence contains two sampling strategies trying to occupy the same chair.
If availability determines who enters your study, you are describing convenience-based recruitment.
Random sampling has a specific meaning.
What is simple random sampling?
In probability sampling, selection is governed by a probability mechanism.
In a simple random sample, units are selected from a defined sampling frame through a random process rather than researcher judgement.
A practical implementation typically requires:
- defining the eligible population;
- constructing or obtaining a suitable sampling frame;
- giving eligible units unique identifiers;
- selecting the required number using a random procedure;
- preserving the selection record;
- recruiting the selected units according to the protocol.
The important part is not that the researcher says the word “random.”
The selection mechanism has to be random.
selecting 200 nurses from 1,200
Suppose your target population is:
All 1,200 eligible nurses employed in 12 health facilities in a district.
You have a complete, current list of all 1,200 eligible nurses.
Your required sample is 200.
A simple random sampling procedure could be:
Step 1: create the sampling frame
Obtain the full list of eligible nurses.
Check it for:
- duplicates;
- people who have left;
- missing facilities;
- ineligible staff;
- missing eligible staff.
Step 2: assign unique IDs
Number the nurses from 1 to 1,200.
Step 3: generate the sample
Use a computer-generated random-selection procedure to choose 200 unique IDs without replacement.
Step 4: preserve the record
Save the selected IDs and the procedure used.
Step 5: recruit according to the protocol
Contact the selected participants.
Do not replace someone simply because:
- they are difficult to contact;
- another nurse is sitting nearby;
- the replacement is more enthusiastic;
- data collection is ending at 5 p.m.
Those decisions can change the sampling process.
Why do I need a sampling frame?
Simple random sampling requires you to know the units from which you are sampling.
That usually means having a reasonably complete list or frame.
If your target population is:
All nurses employed by 12 named facilities
a facility staffing list may make simple random sampling feasible.
If your target population is:
All women aged 15-49 living across a large district
you may not have a current list containing every eligible woman.
Now simple random sampling of individuals becomes much harder operationally.
A multistage design may be more practical.
You cannot randomly sample people from a list they never had a chance to appear on.
Random selection does not repair an incomplete sampling frame.
Is picking every fifth person simple random sampling?
Usually no.
If you select a random starting point and then choose every kth unit from an ordered sampling frame, you are generally using systematic sampling.
Systematic sampling can be a legitimate probability-sampling method.
It is simply not simple random sampling.
Similarly:
- selecting villages first and households second is a multistage design;
- randomly selecting schools and then students within schools creates clustered selection;
- intentionally sampling equal numbers from predefined strata is stratified sampling.
“Random” is a family characteristic shared by several probability methods.
It does not make them all simple random samples.
why district household surveys often need more than simple random sampling
Suppose you want to estimate contraceptive use among women across an entire district.
A theoretical simple random sample of individual women would require a usable frame of all eligible women.
You may not have one.
A more practical probability design could look like:
District
↓
randomly selected enumeration areas or villages
↓
randomly selected households
↓
one eligible participant selected according to a defined procedure
That can still be probability based.
But it is no longer a simple random sample of individuals from the entire district.
That difference affects:
- sample-size planning;
- weighting;
- variance estimation;
- analysis;
- how you describe the design.
Do not call a multistage cluster design “simple random sampling” because random numbers appeared somewhere inside it.
Bench Check: “we randomly selected whoever was available”
You wrote:
“Participants were randomly selected from those available at the clinic on the day of data collection.”
Ask what actually happened.
If the data collector approached whichever eligible participants were present and willing, that is not simple random sampling from the broader clinic population.
If there was a complete list of that day's eligible attendees and a genuine random procedure selected participants from that list, the description may be different.
The sampling label should follow the procedure.
Not the other way around.
Does random sampling guarantee a representative sample?
No.
This is another common overstatement.
Probability sampling provides a principled basis for selection and statistical inference.
It does not guarantee that every realised sample will perfectly reproduce every characteristic of the population.
Random samples are still subject to sampling variability.
And other problems remain possible:
- non-response;
- coverage error;
- outdated frames;
- measurement error;
- differential participation.
A random sample from a bad frame is still a sample from a bad frame.
What happens when selected participants do not respond?
Do not quietly substitute the next available person unless the protocol includes a defensible replacement process.
Non-response is information about the study process.
You should plan for it during sample-size calculation and document it during recruitment.
Depending on the study, you may need to report:
- numbers selected;
- numbers contacted;
- numbers eligible;
- numbers participating;
- reasons for non-participation where available;
- response rate;
- whether responders and non-responders differ on available characteristics.
Probability selection and response are separate stages.
How do I generate a simple random sample?
You can use reproducible random-number tools in statistical software, spreadsheets or programming languages.
The exact tool matters less than documenting:
- the population/frame used;
- the number of units;
- whether sampling was with or without replacement;
- the number selected;
- any stratification or restrictions;
- how the random selection was performed.
Where reproducibility matters, preserve the seed or saved selection output where your software permits it.
What do I actually write in my methods section?
Weak:
Participants were selected randomly.
Stronger:
A sampling frame of 1,200 eligible nurses was obtained from facility human-resource records and each nurse was assigned a unique identification number. A computer-generated simple random sample of 200 unique IDs was selected without replacement. Selected nurses were then approached for participation according to the study recruitment procedure.
If you used systematic, stratified or multistage sampling, describe that instead.
The methods section should allow another researcher to understand what you did.
“Random” is not enough detail.
Look at your sampling paragraph.
Circle every use of:
- random;
- randomly;
- representative;
- selected.
Then ask:
- What was the sampling frame?
- Who exactly had a chance of selection?
- What random procedure was used?
- Did everyone have the selection probability your design claims?
- What happened when selected participants did not respond?
- Does the analysis reflect the actual sampling design?
If you cannot answer those questions, the sampling label may be ahead of the method.
Frequently asked questions
- What is a simple random sample?
It is a sample selected from a defined sampling frame using a random mechanism under a simple random sampling design.
- Is convenience sampling random if I recruit people in no particular order?
No. Lack of a deliberate order is not the same as probability selection.
- Is systematic sampling the same as simple random sampling?
No. Systematic sampling generally uses a random starting point followed by selection at a fixed interval. It can be probability based but is a distinct design.
- Can I use Excel or statistical software to select a random sample?
Yes, provided the frame is appropriate and the procedure is implemented and documented correctly.
- Does simple random sampling eliminate bias?
It can reduce investigator-driven selection bias compared with discretionary recruitment, but it does not eliminate problems such as frame coverage error, non-response or measurement bias.
If your objective is population estimation, read Probability vs Non-Probability Sampling: Which One Fits Your Study?
Then use the Methods Bench Sample Size Calculator only after confirming that the calculation assumptions fit the sampling design you actually plan to use.
References and further reading
- 1.World Health Organization. WHO STEPS Surveillance Manual: Preparing the Sample. https://cdn.who.int/media/docs/default-source/ncds/ncd-surveillance/steps/part2-section2.pdf
- 2.World Health Organization. Survey sampling documentation and health survey resources. https://apps.who.int/healthinfo/systems/surveydata/
- 3.CDC/NCHS. NHANES survey design overview, including sample design and non-response. https://wwwn.cdc.gov/nchs/nhanes/continuousnhanes/overviewbrief.aspx
- 4.CDC/NCHS. National Health Interview Survey Methods. https://www.cdc.gov/nchs/nhis/about/nhis-methods.html
Take it further
Better research, once a week.
Practical methods guidance, research tools and funding opportunities from Methods Bench.
Get practical research notes and new opportunities
Short, useful emails for researchers in Uganda and East Africa. No spam.