Experimental Design
Introduced| English |
|---|
| treatments/ˈtriːtmənts/ |
| experimental units/ekˌsperɪˈmentl ˈjuːnɪts/ |
| placebo/pləˈsiːbəʊ/ |
| factor/ˈfæktə/ |
| control/kənˈtrəʊl/ |
| randomization/ˌrændəmaɪˈzeɪʃn/ |
| replication/ˌreplɪˈkeɪʃn/ |
| control group/kənˈtrəʊl ɡruːp/ |
| blinding/ˈblaɪndɪŋ/ |
The parts of an experiment
- An experiment imposes treatments on units and measures the result.
- Experimental units 实验单元: who or what receives a treatment (people = subjects).
- Treatments 处理: the specific conditions being compared (drug vs. placebo).
- The factor 因子 is the explanatory variable you control; the response is what you measure.
Control, randomize, replicate
- Control 控制: keep other conditions the same so only the treatment differs.
- Randomization 随机化: assign units to treatments by chance — it balances out unknown differences.
- Replication 重复: use enough units (and repeat) so results aren't a fluke.
- These three principles are the backbone of every good experiment.
Control group and placebo
- A control group 对照组 gets no active treatment — a baseline to compare against.
- A placebo 安慰剂 is a fake treatment (sugar pill) that looks real.
- It separates the treatment's true effect from the placebo effect (improving just from believing).
- Comparing treatment vs. placebo isolates what the treatment actually did.
Blinding
- Blinding 盲法 hides who got which treatment, to stop expectations from biasing results.
- Single-blind: the subjects don't know their treatment.
- Double-blind: neither subjects nor the people measuring them know — the strongest guard.
- Blinding keeps hopes and hunches from leaking into the data.
Keep the three principles distinct. Control removes known outside differences; randomization balances the unknown ones; replication (enough units) ensures the effect is real, not chance. Randomization is what makes a cause-and-effect conclusion possible — control and replication alone, without it, are not enough.
Testing a new headache drug.
- Units/subjects: the patients. Treatments: new drug vs. placebo.
- Control group: patients on the placebo. Response: hours until pain relief.
- Double-blind: neither patients nor nurses know who got the real drug → unbiased measurement.
An experiment has experimental units, treatments, a factor (explanatory) and a response. Good design uses control, randomization, and replication; a control group (often on a placebo) is the baseline, and blinding (single- or double-blind) stops expectations from biasing results.
Splitting units into treatment and control
Randomly assign units to each treatment (and a control).
Match each principle to what it does.
Control = known, randomization = unknown, replication = enough units.
A fake treatment (like a sugar pill) that looks real is a...
A placebo controls for the placebo effect.
Neither the subjects nor the people measuring them know who got the real treatment. This is...
Both sides blinded = double-blind.
The group that receives no active treatment, used as a baseline, is the ___ group.
The control group is the comparison baseline.
Randomly assigning units to treatments is what allows a cause-and-effect conclusion.
Random assignment balances confounders, enabling causal claims.