45 terms

experimental research design

research design
the outline, plan, or strategy used to answer the research question; it specifies how to collect and analyze the data
purpose of research design
control for unwanted variation, suggest how data will be statistically analyzed, suggest the conclusions that can be made
weak experimental design
designs that do not control for many extraneous variable and provide weak evidence of cause and effect
One-group posttest-only design
design in which a single group of participants is measured on a DV after having undergone an experimental treatment
why one-group posttest-only design is not useful
no pretest and no control group, w/out a pretest, it is diff to know if the treatment effect produced a change, almost all threats to internal validity apply (useful only when specific background info exists)
one-group pretest-posttest design
single group of participants measures on DV is then administered the treatment and DV is measured again (diff btwn pre and posttest taken as indication of the effectiveness of the treatment)
threats to internal validity of one-group pretest-postest design
history and maturation, (to infer causality must identify and demonstrate that internal validity threats DNE
Posttest-only design w/ nonequiv groups
one group of participants receives treatment and is then compared on the DV to a group of participants that did NOT receive the treatment condition
problem w/ posttest-only design w/ nonequiv groups
is doesnt exclude the possibility of selection effects (participants in comparison group might differ in important ways from the participants int he experimental group
weak designs (3)
1. one-group posttest-only group
2. one-group pretest-posttest design
3. posttest-only design w/ nonequivalent groups
Why are the considered weak designs
bc they do not provide a way of isolating the effect of the treatment, and rival hypotheses are not excluded
Strong designs
provide excellent control for threats to internal validity. so it controls for the effect of extraneous variables via control techniques OR a control group
control group
group that does NOT get the IV or gets some standard value--responses of this group must stand for responses that particpants in teh experimental group would have given if they did not receive the treatment condition
experimental group
group of participants that receives the treatment condition that is intended to produce an effect (group that gets some amount of the IV
what the experimental group participants' responses would have been if they had NOT received the treatment
Two main types of strong experimental designs
1. between-participants design
2. within-participants design
btwn-participants desgin
groups produced by random assignment, and diff gorups are exposed to diff levels of the IV
two types of btwn-participants designs
1.posttest-only control-group design
2. pretest-posttest control-group design
within participants design
all participants receive all levels of the IV (AKA "repeated measures design)
type of within participants design
within-participants posttest-only design
posttest-only control-group design
participants randomly assigned to a treatment condition; the DV is measured only once after treatment condition is administered to the experimental gorup--then responses of experimental group compared to the responses of the control group
pretest-posttest control group design steps:
1. participants are randomly assigned to groups and then pretested on the DV
2. the IV is then administered to the experimental group
3. the experimental and control groups are posttested on the DV
4. diffs btwn pre and post test scores for experimental vs. control group then statistically tested to assess effect of IV
pretest-posttest control group design steps
does great job at controlling rival hypotheses (history, maturation, instrumentation, regression, selection)
btwn-participants design
research participants are randomly assigned to the experimental and control groups (provides the necessary equivalence of groups by randomly assignment particpants)
2 difficulties w/ btwn-participants design
1. randomization does not provide 100% assurance of group equiv
2. is not the most sensitive design fro detecting effect caused by the IV
main function of the pretest
allows researcher to directly observe change in DV as a result of the treatment effect
reason for a pretest design
1. to determine if randomization worked
2. ANCOVA (anal of covariance) to control for any pretest diffs
3. increase sensititvity
4. test for ceiling effect
to test for initial position of participants on the DV
5. to obtain evidence of change
drawbacks to pretesting
1. may increase amt of time/money req'd to complete study
2. may sensitize participants to the experimental treatment condition (most serious)
why is sensitization of participants to treatment condition most serious prob?
1.participants knowledge could heighten their sensitivity to the IV
2. participants responses may produce responses that are NOT representative of those that would have been obtained if pretest not occured
3.results might not generalize to other participants who have not taken pretest
4. effet of pretesting is dependent on the type of study conducted
within-participants posttest-only design
same research participants are repeatedly assessed on the DV after participating in all treatmnet conditions (so the same 20 participants in each treatment condition)
advantages of within-participants design
equiv of research participants 100% bc they are the same participants involved in each treatment condtion (so participants serve as their own control, variables remain constant over entire experiment, and increases sensitivity of experiment-maximally sensitive to the effects of the IV) & requires fewer participants than BTWN-participants design
main disadvantage of within-participants design
sequencing effect bc same participants participating in more than 1 treatment condition (counterbalancing can help) & cannot control for threats of history, maturation and regression to the artifact
Factorial design
two or more IVs are simultaneously studied to determine their independent and interactive effects on the DV
how are factorial designs made?
1. participants randomly assigned to groups
2. a factorial design is described in terms of the number of independent variables it uses and the # of conditions or levels of each IV
3. a 2x2 design has 2 IVs each w/ 2 conditions
4. a 3x3 design has 3 IVs two of which has 3 levels and one that has 2 levels
components of a factorial design are?
1. cell
2. main effect
3. interaction effect
treatment combo of 2 or more IVs
main effect
effect (ie sep influence) of one IV has on the DV (so doesn't involve the other IV)
interaction effect
it occurs when 2 or more IVs combine to produce an effect over and above their main effects
factorial design based on a mixed model
factorial design that represents a combo of the within-participants and the btwn participants designs
steps in factorial design based on a mixed model
1. participants are randomly assigned to diff levels of variation of the btwn-participants IV
2. all participants then take each level of variation of the within participants variable
main characteristic of factorial design based on a mixed model
has both a btwn and within component, so at least one IV requires diff participants for each level of variation, and at least one IV req's the same partipants in each level of variation
advantages of factorial design based on a mixed model
can test for effects of each of the two IVs (btwn) and need fewer participants bc all participants take all levels of variation of one of the IVs within
disadvantages of factorial designs
increase the # of research participants, when maninpulating more than one IV, increased difficulty of doing so simulataneously, and diff in interpreting higher order interactions (like a 3way)
advantages of factorial designs
1. can manip more than one IV, which allows us to test multiple/more precise hypoth at once
2. can control potential confounding extraneous variable by building it into the design (gender)
3. provides greater precision when you add mor than one IV
4. can test the effect of interactions, which can give us a better understanding of the given phenomenon
typically which design is more favored?
within participants due to increased sensitivity of IV effect