Experimental Design and Analysis Exam #2

John Carroll University- Experiemental Design and Analysis with Dr. Yost. Exam #2 covers chapters 8-10. The cards are mostly based on Yost's chapter outlines using supplementary definition from the text.

confounding variable

component reasonably responsible for observed relationship not measured in the study

how to work around confounds in research

goal is to control for confounds through random assignment (creates equivalence in the IV groups)

confounds in relation to validity

controlling for confounds increases a study’s internal validity

Posttest only design

advantages of pretest-posttest design

1. determining if there is equivalence prior to manipulation

2. good if you need to identify a particular type of participant (ex. - maybe in a “quit-smoking” study, heavy smokers saw less progress with the treatment method than light smokers)

disadvantages of pretest-posttest design

1. time consuming and awkward

2. Can sensitize participants to the hypothesis that could lead to response bias

3. might effect generalizability as pretests do not really occur in the real world

Independent group/ between subject design

Ps only exposed to 1 level of the IV

Repeated Measure / within subject design

Ps exposed to both IVs/treatments

advantages of repeated measure design

1. fewer Ps needed

2. increased statistical significance because of deceased random error

disadvantages of repeated measure design

1. order effect

2. practice/learning effect

3. fatigue effect

4. carryover effect

order effect

order in which treatments are presented affects the DV; includes contrast effect

practice/learning effect

performance on task 2 only improves because of increased familiarity/practice gained in task 1 (could be tempered through counterbalancing)

fatigue effect

opposite of practice effect; performance on task 2 is low from boredom, tiredness, etc from task 1

carryover effect

effects from treatment 1 affects performance in treatment 2; often deals with the time between treatments

Factorials

Notated by a “!”; the number that precedes the “1” multiplied by all numbers that come before it numerically

2! = (2×1)= 2

3! = (3×2×1)= 6

4! = (4×3×2×1)= 24

matched pairs design

aims to match people based on a participant variable; goal is to achieve the same equivalence of groups as repeated measures without the need to put ps through both treatments

Analysis of Covariance (ANCOVA)

allows for statistical equivalence of differences (ex- if the goal is to measure who can jump higher, you would need to statistically make ps the same height to get a true jump height measurement)

selecting research participants

if goal is to describe the population, you must use probability sampling

if goal is to test for a relationship between variables, nonprobability sampling can be used

straightforward manipulations

presenting written, verbal, or visual material to ps (ex- for memory and recall studies)

staged manipulations

involved acting, fake scenarios, deception; used to properly manipulate the variable (ex- doing a study to measure the effectiveness of assertiveness training but the real test was whether or not the ps asked for the pay they were promised after being shortchanged)

Using strong manipulations

Generally researchers want to use the strongest possible manipulation, especially in early stages when they are trying to support the existence of a relationship between variables. HOWEVER using the strongest possible manipulation is tempered by ethics (i.e a strong manipulation of fear or anxiety) and a potential decrease in external validity (the stronger a manipulation, the less likely it is to occur irl, thus decreasing its ability to applied to larger populations or daily life)

types of DV measures

1. self reported measures

2. behavioral measures- direct observation of behaviors

3. physiological measures such as GSR (measures anxirty through sweating and skin electricity), EMG (measures muscle tension), and EEG (measures brain cell activity)

4. behavioroid measures- measures behavioral intent

sensitivity of the DV

ceiling effect and floor effect

ceiling effect

task is so easy everyone does well, leaving little room for improvement

floor effect

task is so difficult everyone performs poorly

controlling for participant expectations

1. placebo effect

2. demand characteristics

placebo effect

ps given a fake treatment to see if the power of mind effects results

demand characteristics

ps figuring out hypothesis and having biased answers

single blind study

ps do not know whether they are part of the treatment or placebo group

double blind study

neither ps nor researchers know which group ps are in

expectancy effect

experimenter bias leading to skewed results

manipulation checks

attempt to directly measure if the IV has the intended effect on the DV (does it actually measure what its supposed to?)

F value

F =between group variability (÷) within group variablility

Goal is to have F value as large as possible

within group variability can be reduced by

1. people’s individual differences

2. setting the stage step of the experiment

Increasing the # of levels of an IV allows…

…us to test for curvilinear relationships (you need at least 3 data points); the more levels you have, the more detailed your data will be

Effects of increasing # of IVs

1. increasing external validity

2. allows for economic use of participants

main effects

does the IV have an effect of the DV regardless of the other variable?

interaction

the effect of one IV on the DV depends on the other IV

simple main effects

compares the mean difference at each level of the variables

how to find # of conditions in an experiment

multiply numbers in format

When graphing data, how can you tell if variables interact with each other?

if when you graph the means of the main effects and connect the data points the lines/patterns are not parallel, there is an interaction

For purposes of this course/exam, what is the indicator of significant results?

If the difference between the two means is equal to or greater than 2

What is an IV x PV design?

any design where there is a manipulated variable (IV) and a participant variable (PV)

Participant variable (PV)

• fixed; ps bring in this condition with them when the partake in the study.

• Examples: self esteem (from class examples), sex/gender

Know how to solve and interpret basic data sets

Example