Stat A Experimental Design Quiz

Replication

Having multiple experimental units / subjects and wanting enough people / things so that cause and effect are not just by chance, which reduces variability and is more precise

Randomization

Random assignment to treatments which spreads out the types of people and variables we can't control, an example being simple random selection

Control

Keep the variables consistent to prevent confounding variables

Confounding Variable

A variable that may alter results

Control Group

Provides a basis of comparison

Placebo

A fake treatment

Placebo Effect

Feeling better just because you got any sort of treatment

Single Blinding

The subjects don't know their treatment

Double Blinding

The subjects and researchers don't know their treatment

Observational Study

No variable manipulation or influence, just watching and observing

Experiment

Impose a treatment to determine cause and effect

Experimental Units

The things or people are participating in an experiment, humans equal subjects

Factor

The independent variable or what we are manipulating to form a treatment

Level

The specific value of a factor

Treatment

The combination of factors, levels, and controls, if there is only one factor, the treatments equal the levels

Response Variable

The dependent variable or what we are measuring and comparing at the end of the experiment

Completely Randomized Design

All experimental units are together and randomly assigned to treatments, similar to a simple random sample

Matched Pairs

A special type of blocking with only two units and treatments per block, keeping each pair of units as similar as possible as your own form of control, which eliminates most confounding variables

Randomized Block Design

Split experimental units into blocks based on a common characteristic that may affect the response or be confounding and then run the experiment within each block, this reduces variation and is similar to stratified sampling