OVERVIEW: There are good techniquesfor producing data. There are also bad techniques that produceworthless data. Random sampling and randomized comparativeexperiments are extremely important and effective statisticalpractices. The use of chance is vital in statistical design.

Observational study... no treatment imposed on individuals. Simply measurevariables of interest.

Experiment...treatment imposed on individuals in order to observeresponses.

Experimental units:Individuals on which experiment is done.

Treatment: Aspecific experimental condition applied to experimental units.

Placebo: A dummytreatment that can have no physical effect.

Control group...receive dummy treatment. (Helps experimenter control effects oflurking variables.)

Completely randomized experimentaldesign: All experimental units areallocated at random among the treatments.

Statistically significantobservation... an observed result toounusual to be an outcome determined by pure chance.

Three principles of experimentaldesign

1. Control... needed to countereffects of lurking variables, etc.
...Simplest form of control is comparison. Experiments shouldcompare two or more treatments in order to avoid confounding theeffect of a treatment with other influence.

2. Randomization...subjectsassigned to treatments by pure chance.
...Creates groups that are similar, except for chancevariation. Table of random digits can be used to choose treatmentgroups.

3. Replication...do experimenton many subjects to reduce chance variation in the results.

Double blind experiment: Neither the subjects nor the people who have contact withthem know which treatment a subject received.

Block design...usedto minimize variation. A block is a group of experimental units orsubjects that are similar in ways that are expected to affect theresponse of the treatments. Treatment is assigned randomly withinsimilar blocks. Blocking are a form of control.

Matched pairs: Acommon form of blocking for comparing two treatments. The two pairsare "alike." Common forms of matched pairs include...

-Using random process. One of pairreceives treatment, the other a placebo. Pairs observed at a latertime to see if treatment had effect.

-Test scores from a before-after situation. Anindividual takes a before-test, then gets some type of treatment,than takes an after-test (similar to the before-test). Purpose is tosee if treatment improves test performance.

RETURN TO TEXTBOOK HOME PAGE /Back to the top of this page