🏥 Thanjavur Medical College (TMC) Case Conference 🩺

1. Set up your survey

Sample size

People interviewed50

How will you select people?

Repeated surveys to run

Separate samples20

What are we estimating?

Teaching village: 1,000 people. This is a fictional population so we can compare your estimate with the known answer.

2. Watch the village and sample

Has condition / did not receive serviceNo condition / received serviceNot selected

Whole village truth16%

Your sample—

Estimate error—

Choose a sampling method and tap Run the survey. First predict: will your sample match the village exactly?

3. See how estimates fluctuate

Each point is a fresh survey using the same method and sample size. The dashed line is the true village value.

Average estimate—

Lowest–highest—

Near truth (±5 points)—

Your field-note card

—

Your latest estimate

Ready to sample

The estimate will appear after you draw a sample.

“BIGGER helps; BETTER selection matters.”

A large biased sample can be confidently wrong. Sampling design comes before sample size.

Field note: A sample is a small window onto a population—not the population itself.

Quick memory hooks

SAMPLE

Select fairly • Ask a clear question • Mix in all groups • Plan size • Look for bias • Explain uncertainty.

“More people, less wobble—unless the doorway is wrong.”

Increasing n usually reduces random fluctuation. It does not remove selection bias.

“Random is not haphazard.”

A probability sample uses a planned method so eligible people have a known chance of selection. Convenience sampling just takes whoever is easiest to reach.

“Clinic ≠ community.”

People attending a clinic may be sicker or more health-conscious than those who do not attend. Clinic data answer clinic questions—not automatically village questions.

What just happened?

Run a survey to get a short interpretation linked to your choices.

Why do estimates change between surveys?

Different random samples include different people. This is sampling variability. With small samples, a few people can move the percentage a lot.

Why can a large survey still be wrong?

If the selection process systematically misses or over-represents a group, the estimate can be biased. A bigger sample may make that biased estimate more precise without making it representative.

What does “±5 percentage points” mean here?

It is a simple visual learning target, not a formal confidence interval. Real uncertainty depends on design, weighting, non-response, clustering and other factors.

Thanjavur Medical College (TMC) Case Conference Group

Village Survey Safari

1. Set up your survey

2. Watch the village and sample

3. See how estimates fluctuate