Statistical Terms

Question 1

Factors

Answer 1

We group related treatments into a single factor

ie: 3 different amounts of fertilizer would be one factor, another would be 4 different amounts of irrigation

Question 2

Treatments

Answer 2

The procedures that we want to evaluate

Question 3

Levels

Answer 3

The treatments that make up a factor are called the levels of the factor (sometimes are ordinal, sometimes nominal)

Question 4

Experimental Units (IMPORTANT!)

Answer 4

(Sometimes called true replicates)

The individual “entities” (subjects, plots, objects, groups, etc), that are assigned to and that receive individual application of a factor level (or combo of factor levels).

(ie: Human subject is assigned to certain pill, a pen full of turkeys is assigned a certain diet [individual turkey is the measurement unit], a class full of students is assigned a certain pedagogical intervention, a single batch of cookies is baked at a certain temperature)

Question 5

Sample Size (or number of true replicates)

Answer 5

This usually refers to the total number of experimental units

Question 6

Response Variable (or Outcome)

Answer 6

This is what we measure on each experimental unit. How do the factor(s) affect these measurements? (often more than one)

ie: 
• Blood pressure in humans
• Mortality rate for young turkeys
• Yield of a crop
• Average score on an exam

Question 7

Measurement Units (or Pseudo-Replicates)

Answer 7

This is the unit that measurements of the response variable are actually made on.

ie:
• Feed a pen of turkeys all the same, and actually measure a random sample of 30 individual turkeys
• Bake a batch of cookies, measure the hardness of each individual cookie.

Question 8

If the measurement unit is not the same as the experimental unit, to get the experimental unit we…

Answer 8

take the average of the measurement units!

Question 9

Control (define for verb and noun)

Answer 9

verb: When an experimenter gets to allocate the experimental units
noun: When new treatments are compared with a standard treatment (or no treatment at all). That standard/no treatment is called the “control”

Question 10

Blind

Answer 10

An experiment involving humans in which the subjects are not told which treatment they received.

If the experiment does not involve human subjects, blind refers to the fact that those administering/evaluating the experiment don’t know which treatment the subject receives.

This avoids (behavioral) confounding factors.

Question 11

Double blind

Answer 11

An experiment involving humans in which the subjects and the people evaluating them ALL do not know which treatment the subject received.

Again, avoids confounding factors.

Question 12

Placebo

Answer 12

A fake treatment that resembles the true treatment.

This helps maintain the “blindness” of an experiment, so that the ONLY difference between groups is the treatment itself.

Question 13

Evaluators should be blinded if…

Answer 13

any subjectivity is possible.

Question 14

Experimental Error (or Chance Error)

Answer 14

This is the difference among repeated identical experiments (not a mistake).

An objective of experimental design is to reduce the size of experimental error as much as possible.

Question 15

Confounding

Answer 15

When the effects of two or more factors cannot be distinguished.

Question 16

How can we perform good allocation of Experimental Units?

Answer 16

Use some probabilistic mechanism (coin, die, SAS program)

Question 17

Why do we randomize?

Answer 17

Randomization is the basis for ALL statistical inference carried out on experimental data

Question 18

Statistical inference

Answer 18

The process of drawing population-level conclusions from sample data. (includes hypothesis testing)

Question 19

An experiment without randomization is a poor experiment and cannot be analyzed statistically because….

Answer 19

of possible confounding factors

Question 20

Sampling Distribution

Answer 20

The theoretical distribution our test statistic would have if our H_0 is true.