between-subject design
= each treatment is administered to different group of subjects
- data collected from subjects within a given treatment are averaged —> any effect of independent variable will appear as difference in these treatments averages
- randomized two-group design
- randomized multi-group design
within-subject design
= a single group of subjects is exposed to different all of the treatments (one treatment at the time)
- data collected from subjects within a given treatment are averaged —> any -effect of independent variable will appear as difference in these treatments averages
single-subject design
= focus on changes in behavior of single subject (or small number of individual subject) under different treatment condition; no average data set across subjects
error variance
= variability among sources caused by variables other than your independent variable
single-factor randomized-group design
- assign subjects at random to different groups –> expose each group to different level
2 variants:
- -> randomized two-group design
- -> randomized multi group design
randomized two-group design
= assign subjects random to 2 groups + expose 2 groups to different levels + hold extraneous variable constant
randomized multi group design
= expand randomized two-group design by adding one or more levels (can include as many levels as needed)
2 ways to manipulate independent variable:
- quantitatively –> using parametric design (systematic variation of the amount of independent variables)
- qualitatively –> nonparametric design
multiple control group design (randomized multi group design)
- used when single group is not adequate to rule out alternative explanations of results
- variation on single-factor multi group design –> one that includes control groups
matched-groups design
= matched sets of subjects distributed at random, one subject per group, into groups of experiment
(when characteristics subjects correlate strongly with dependent variable –> asses subjects on one/ more characteristics & dived them equal into different treatment groups)
matched-pairs design
= patched group equivalent to randomized two-group design
confounding
- biased sample: select random sample
- biased experimenter: blind study
- extraneous variable( which changes
with independent variable): careful planning of how confounding could play a role
reducing error variance
- holding extraneous variables as constant as possible (eg experimental protocol) - match subjects on influential characteristics (see Matched-Group or Within-Group Design)
