Short answer
Random sampling determines who enters a study; random assignment determines which treatment group an enrolled participant enters. Sampling concerns how well the study can represent a larger population, while assignment concerns the credibility of comparisons between treatment groups. One does not establish the other. 1
On this page
At a glance
Scroll to compare all columns →
| Question or attribute | Random sampling | Random assignment |
|---|---|---|
| What is chosen? | Units included in the sample | Treatment or control allocation |
| Starting pool | A population or sampling list | Study participants |
| Main concern | Generalizability | Treatment comparisons |
| Validity connection | External validity | Internal validity |
These are the distinct roles described in the clinical-research overview. 1
What each thing is
Sampling moves from a larger population to a smaller study sample. Simple random sampling gives each subject an equal selection chance; stratified sampling instead divides the population into subgroups and draws samples within them. Assignment operates on participants selected for the study, distributing them among treatment conditions through chance. 1
Key differences
The two processes address different sources of uncertainty. Sampling supports inference from the studied people to a larger population. Assignment helps separate treatment differences from differences between the people receiving those treatments. A strong treatment comparison therefore does not, by itself, establish that the participants represent everyone to whom the results might be applied. 1
How to tell them apart
Ask what the random procedure decides: inclusion in the study, or allocation within it? A random draw from a census list indicates sampling; a random sequence assigning enrolled patients to treatment or control indicates assignment. The limit: the word “randomized” alone is insufficient, because the supplied sources do not use all related terms consistently. 1 2
Where they overlap
Both use chance, and both can use computer-generated random numbers. A study could hypothetically select participants randomly and then randomly allocate those participants to treatments. These remain two separate steps: the first concerns the study’s population connection, and the second concerns its comparison groups. 1
Edge cases
An eligible, consenting patient sample can receive random treatment assignment without having been randomly sampled from a broader population. The treatment comparison may benefit from randomization while broader applicability remains a separate question. Likewise, dividing people into strata is not enough to identify the process: check whether the subsequent random choice selects participants or allocates treatments. 1 2
Why the distinction exists
Researchers ask both “Who do these findings describe?” and “What explains the difference between these groups?” Keeping selection and allocation separate prevents evidence addressing one question from being treated as an answer to the other. Study setting also matters: the first source notes that hospital and population-based samples may show different disease profiles. 1
Common misconceptions
Random sampling does not mean every realized sample perfectly mirrors its population; the source recognizes random sampling error. Random assignment is not a certificate of population representativeness. Nor should it be read as guaranteed exact baseline equality: the second source calls for examining group characteristics after randomization. 1 2
Examples
-
In the Pakistan prevalence study described by s1, investigators randomly chose census-listed households and interviewed their members. That is sampling: the random choice determined whose data entered the estimate. 1
-
In the meningitis trial described by s1, consenting participants received dexamethasone or placebo through random allocation. That is assignment: the random choice determined treatment, not selection from the wider population. 1