Vocabulary

Experiment -- an investigation to determine the effects of a explanatory variable on a response variable

Response variable -- a variable that is measured based on the outcome of a study (think of it as a dependent variable or Y)

Explanatory variable -- explains the observed effect on a response variable (the treatment)

Experimental Units -- subjects of an experiment

Factor -- a variable controlled by the experimenter (another word for explanatory variable)

Level - a specific value or type of factor

Extraneous factor -- a variable that may influence the response variable

Confounding variable -- a variable that isn't monitored but is influencing our results

For example, let's say a professor compares two different teaching styles -- engaging and non-engaging on his groups of students. He or she implements an engaging style in the spring and the non-engaging style in the fall. The students naturally preferred the engaging style. But a confounding variable in this study might have been the seasons themselves. Students may be sluggish in the fall and more happier in the spring. The confounding variable (seasons) makes it hard to determine if seasons or the teaching style may have influenced the study.

Blocking -- the experimental equivalent of stratifying; make "blocks" (aka groups) of experimental units with similar traits. Blocking is primarily used to minimize the effects of the variability in the experimental units.

Control group -- a benchmark group used to compare the treatment group

Treatment -- what the experimental units do in the study (can be taking a test, running a mile, etc.)

Observational study -- a study that looks at data to observe a characteristic; can't be used to determine a cause-effect relationship between variables, just an association

Blinding -- the experimental units don't know what treatment they are given

Double-blinding -- experimental units and any researcher/evaluator of the data both don't know what treatment the units have been given

Placebo -- a fake treatment that has no effect; used as a benchmark or control group

Replication -- a key component in experimental design; refers to how an experiment must have numerous subjects/units involved (not just one subject)

Control, randomization, and replication are all key parts in an experimental study!

Statistically significant - values/effects that typically show up so commonly that they aren't due to chance or luck

How to Answer Experimental Design Questions

Question 1: A dog food company wants to see if their new food lineup is improving the health of its dogs compared to its old food lineup. The company selected 40 volunteered dogs to participate in the experiment. Design an experiment for the company.

Make a flow chart! You don't have to write a paragraph describing the entire process!

The only part you have to explain is random assignment:

Number the dogs from 1 to 40 with no repeating numbers. Have an RNG generate 20 unique numbers from 1 to 40. The dogs whose numbers were generated go to the new food lineup treatment group. The rest go the old food treatment group.

But what about blocking? Let's say the company thinks that dog size (large vs small) influences the results. There are 20 big dogs and 20 small dogs. Design an experiment that blocks based on size.