5.2 power point

Upload: krothroc

Post on 08-Apr-2018

217 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/7/2019 5.2 Power Point

    1/19

    Designing ExperimentsSection 5.2

  • 8/7/2019 5.2 Power Point

    2/19

    Vocabulary of Experiments

  • 8/7/2019 5.2 Power Point

    3/19

    More

    Vocabulary

    Placebo: a dummy treatment

    Placebo Effect: when a subjectresponds favorably to a placebo

    Control Group: a group that provides astandard for comparison to evaluatethe effectiveness of a treatment; oftengiven the placebo.

  • 8/7/2019 5.2 Power Point

    4/19

    Blinding

    Single-blind experiments: the subjectsdo not know which treatment (or

    placebo) the are receiving.

    Double-blind: if both the subjects andthe researchers dont know which

    treatment each subject receives.

  • 8/7/2019 5.2 Power Point

    5/19

    The Three Key Principles:

    The three key principles ofexperimental design are:

    Control

    Replication

    Randomization

  • 8/7/2019 5.2 Power Point

    6/19

    CRR

  • 8/7/2019 5.2 Power Point

    7/19

    Whos Really in Control?

    Dont confuse Control with ControlGroup.

    Control refers to the overall effort to

    minimize variability in the way theexperimental units are obtained and

    treated.

  • 8/7/2019 5.2 Power Point

    8/19

    The Heart of the Matter We carry out experiments to see if there is a

    difference in the responses of those whoreceived the treatment and those who didnt.

    We hope to see a difference in the responses

    so large that it is unlikely just to happen bychance variation.

    We can use the laws of probability to learn ifthe differences in treatment effects are larger

    than we would expect to see if only chancewere operating.

    If they are, we call them statisticallysignificant.

  • 8/7/2019 5.2 Power Point

    9/19

    Statistical Significance

  • 8/7/2019 5.2 Power Point

    10/19

    An Aspirin a Day

    You often see the phrase statisticallysignificant in reports of investigations inmany fields of study. That means that

    investigators found good evidence for theeffect they were seeing.

    For example, the Physicians Health Studyreported statistically significant evidencethat aspirin reduces the number of heartattacks compared to a placebo.

  • 8/7/2019 5.2 Power Point

    11/19

    Completely Randomized Design

    (CRD)

    When all experimental units are randomlyallocated among all the treatments (ortreatments to the units), it is said to be acompletely randomized design.

    CRDs can compare any number oftreatments. The treatments can be formedby levels of a single factor or by more thanone factor.

  • 8/7/2019 5.2 Power Point

    12/19

    You Blockhead! Suppose a fitness instructor believes that a certain

    exercise regimen will increase upper-body strength.He recruits students to test his theory by havingthem do as many push-ups as they can after they

    complete the training. We would expect some variability due to natural

    differences in strength. We try to control for theseinherent differences by placing subjects into groupsof similar individuals. Because we know that women

    have less upper-body strength than men, we wouldput them into different groups, or blocks.

    By separating subjects by gender, we can reducethe effect of variation in strength on the number ofpush-ups.

  • 8/7/2019 5.2 Power Point

    13/19

    Randomized Block Design (RBD)

    Ablockis a group of experimental units thatare known to be similar in some way that isexpected to systematically affect the

    response to the treatment.

    In a randomized block design, the random

    assignment of units to treatments (ortreatments to units) is carried out separatelywithin each block.

  • 8/7/2019 5.2 Power Point

    14/19

    The Mantra

    Blocking allows us to draw separate conclusions abouteach block. Blocking also allows more precise overallconclusions because the systematic differencesbetween the blocks can be removed when we study

    the overall effects of the treatments.

    A good experimenter will form blocks based on themost important, but unavoidable, sources of variabilityamong the experimental units. Randomization will then

    average out the effects of the remaining variation andallow an unbiased comparison of the treatments.

    The mantra: Control what you can, block on what youcant control, and randomize the rest.

  • 8/7/2019 5.2 Power Point

    15/19

    Can You Elaborate?

    CRDs are the simplest experimentaldesigns, but they are often inferior tomore elaborate designs. In particular,

    matching subjects in various ways canproduce more precise results thansimple randomization.

    The simplest form of matching is calleda matched-pairs design.

  • 8/7/2019 5.2 Power Point

    16/19

    ABetterComparison? To compare two treatments (or a treatment

    and a placebo), you can use experimentalunits that are as alike as possible. The idea isthat matched units are more similar thanunmatched units, so comparing responses ina number of pairs is more efficient thatcomparing the responses of a group ofrandomly assigned subjects.

    Randomization is still important though! Theorder in which the treatment is imposed (orthe assigning of a treatment and a placeboto each unit in a pair) should still berandomly generated.

  • 8/7/2019 5.2 Power Point

    17/19

    Caution!

    Remember not to generalize!

    Statistical analysis of an experiment

    cannot tell us how far the results will

    generalize to other setting orindividuals.

  • 8/7/2019 5.2 Power Point

    18/19

    Outlines of Experiments

  • 8/7/2019 5.2 Power Point

    19/19

    Jasper Leigh