small-n designs and solving problems (control issues) within subjects designs (small n) n stands for...

17
Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs we have discussed in the course so far have had many participants per group In small-N designs you can do experiments with very few participants in fact you can do studies with only 1 participant.

Upload: warren-nichols

Post on 23-Dec-2015

243 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

Small-N designs and Solving problems (control issues)

• Within Subjects Designs (Small N)• N stands for the number of subjects in an

experiment• The designs we have discussed in the course so

far have had many participants per group• In small-N designs you can do experiments with

very few participants– in fact you can do studies with only 1 participant.

Page 2: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• B.F. Skinner was an advocate of small-N designs.– in fact he was a critic of large-N designs– Skinner advocated eliminating as much error

variance as possible

• Two main sources of error variance– 1) that do to individual differences among

participants– 2) That due to ineffective control procedures

Page 3: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• IF you only use one subject in your experiment individual differences will be eliminated as a source of error variance

• If you have highly controlled experimental procedures that will help with physical extraneous variables– Operant chamber

• no light enters• sound attenuated

Page 4: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• Large N designs can obscure actual effects– rare steak Water

• Who uses Small–N designs– clinical psychopathology

• small #s of patients• can ensure everyone gets treatment

– Animal researchers

Page 5: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• Types of Small N designs– The ABA design

• A baseline• B treatment• A return to baseline

– Rat bar pressing• 1 hour no Rf• 1 hour Rf• 1 hour no Rf

Page 6: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

ABABAAB1 AB2 AB3 A

ABABexperimentally sound and ethicalHorne (1987)

facial screening mentally impaired 8-year old girl

spoon bangingsoft cloth over face

ABself injurious behavior

Multiple BaselineVanbieauliet, Spangler, and Marshall (1981)family style meals and communication in instututionalized adults

Small Nuseful when limited # of subjectsHelpful, because can leave patient in a treated state ABABExternal Generalizability can be an issue.

worked for this patient, but would it work for other patients.

Page 7: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• Solving Problems: Controlling for extraneous variables• When we run an experiment we want conditions to be such that only

the IV should change systematically from one condition to another.• We have already discussed how to control for subject variables

– random assignment– matching subjects– within S’s designs

• But there are other types of extraneous variables that must be controlled as well– Physical variables

• aspects of the experimental environment– Social variables

• refers to the relationship between experimenter and participant• We will talk about how to control each of these.

Page 8: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• Physical variables– These are aspects of the testing conditions that need to be

controlled.– Any time you test one group in one setting and another group in

a different setting you have the potential for confounding of physical variables.

• Skinner developed the operant chamber or Skinner box in an attempt to control physical variables.

• Show them an operant chamber• Things to notice

– sound attenuated– masking noise– outside light does not enter– timing of events is carefully controlled by a computer.

Page 9: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• Techniques to control for extraneous physical variables.

• Elimination – this is just what it sounds like. It is an attempt to make some physical extraneous variable go away. By making it go away it cannot have any effect on your experiment.– The operant chamber eliminates unwanted outside

noise and light.– We also try to run human experiments in quiet places

where extraneous physical variables can be eliminated as well.

Page 10: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• Constancy – If we cannot eliminate a physical variable then we try to keep it constant across the conditions of the experiment.– Typically people test pigeons and rats in the same operant chamber.

Therefore the physical characteristics that cannot be eliminated, are constant for all animals.

– The dimensions of the box, lights that do exist, sound of the white noise• For human subjects similar considerations should be made. Try to

test all of your subjects in the same room.– There is some evidence that even things as subtle as color of a room

can have effects, so if possible you should try to keep these things constant.

• Especially important for human subjects is to keep instructions constant for all subjects.

• it would be best if instructions were written out– or video or audio taped

Page 11: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• Sometimes we are unable to keep eliminate or keep extraneous physical variables constant.

• In that case we must make sure that the variable is balanced.– It is OK to use two different rooms to test subjects in, if we have to.– make sure that both groups are equally represented in both rooms.

• Why is this important?– This is called balancing. We are distributing the effects of an extraneous physical variable

across the treatment conditions.– It may increase the variability in responding, but it should not be able to cause the effect.

• How do we go about balancing extraneous physical variables?– first randomly assign half of your subjects to room 1, and the other half to room 2.– Then randomly assign half of the subjects in room 1 to group1 and the other half to group2. – Do the same for the subjects in room 2.

• Balancing is also used to control for issues like time of day of testing.– In an operant chamber you can only test one subject at a time, so you need to make sure

that your different treatment groups are distributed throughout the running day.– run some experimentals and controls in the morning, and some of each in the afternoon.

Page 12: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• Social Variables are another type of extraneous variable that needs to be controlled.– This refers to the relationships between

subjects and experimenters that could influence results.

• There are two principle social variables– Demand characteristics– Experimenter bias

Page 13: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• Demand Characteristics.– aspects of the experimental situation that demand that people behave in a certain way.– An experiment can encourage subjects to try to be Good Subjects.– Subjects may attempt to guess the experimental hypothesis and adjust their responses to

support the hypothesis they think is being tested.• The Hawthorne effect.

– Industrial Organizational psychologists tried to study environmental factors on employee productivity.

– They tried changing lighting conditions, rest periods, temperature, wage rate, and so on.• The interesting finding was that no matter what they changed, performance

increased.– decrease in light increased productivity, just as much as an increase.

• Why would this be?

• It would seem that the workers were pleased that the company was taking interest in them and were honored to be subjects, and thus there performance would increase.

– Increase in performance, because you know you are being studied, but not because of a manipulated variable is an artifact.

Page 14: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• Martin Orne became interested in artifacts caused by subjects trying to conform to beliefs they had about experiments.

– His first interest in this area came from his research with hypnosis.– He had become convinced that some of the trance like behaviors exhibited by people that

were under hypnosis were caused by expectancies the subjects had.– In other words they acted in the way that they thought they should based on common

knowledge passed on about hypnosis.• To test this Orne ran an experiment.

– Group 1: a group of college students were given a lecture on hypnosis. As a part of this lecture they were told about some of the common characteristics of being under hypnosis.

• one of the these characteristics was made up• Catalepsy of the dominant hand.• This was presented as a classical characteristic, and a big deal was made of how it was always the

dominant hand that became rigid.• There was a fake demonstration of hypnosis that showed this phenomenon.

– Group2: a group of college students were given the same lecture and demonstration, but they were not told the made up part about dominant hand catalapsy.

• A few weeks later students from both lectures were asked to serve as subjects in a hypnosis study.

– They found that almost all subjects in Group1 had dominant hand catalepsy under hypnosis, whereas none in Group2 did.

Page 15: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• Orne, who was interested in this finding wondered how compliant people under hypnosis might be– He tried to devise a task that no sane person would be willing to

do.– In his pilot work he brought subjects in that were not under

hypnosis and asked them to add together random paired numbers.

• There were 224 pairs of numbers to add for each sheets and subjects were placed in a room with 2000 sheets.– The experimenter assigned them the task and said he would

come back later.– 5 and a half hours later the experimenter gave up.

Page 16: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• Orne & Scheibe (1964)– sensory deprivation

• group 1 control group• group 2

– emergency tray– alarm button– log book to report unusual experiences

• How to control for demand characteristics– Single Blind Study

• subjects don’t know which group they are in• cranberry juice, mint oil, with or without alcohol

– keep in mind many drugs produce side effects» placebos that have peripheral effects

– Cover story• Nisbett – culture of honor – “asshole” study

Page 17: Small-N designs and Solving problems (control issues) Within Subjects Designs (Small N) N stands for the number of subjects in an experiment The designs

• Experimenter Bias– When the experimenter’s expectations

inadvertently effect the outcome of the study– Clever Hans– Rosenthal

• Intellectual Bloomers• Maze bright; Maze Dull

– Robert True – Age Regression– Double Blind Studies

• require two experimenters