Module 11: Complex Experimental Designs Flashcards
Basic experimental designs
Research design featuring one independent and one dependent variable
Complex designs
experiments using more than one variable (either independent ot more than 1 dependent variable)
Factorial design
a type of complex design with more than 1 independent variable
Multivariate design
more than one dependent variable
What is the 2 option if you want to investigate two independent variables?
- conduct separate experiments
- manipulate each independent variable of interest or manipulate 2 independent variables in the same experiment
What us the benefit of combining two seperate experiments into one?
More individual tests of significance one runs, the greater the likelihood of finding false positive results
2X2 factorial design (3)
- Includes 2 independent variables and 1 dependent variable
- You have 2 independent variables and each has 2 levels
- results in 4 treatment groups
3 X 2 factorial design (2)
- you have 2 independent variables with one variable having 3 levels and the other 2 levels
- Results in 6 treatment groups
Small sample sizes—– the power of a study
lower
confirmattory hypothesis
The researcher specify what they expect to find
Ex: “Playing adventure type games will result in higher conigtive functioning”
Exploratory Hypothesis
The researcher is not interested in confirming a prediction but only in investigating how two variables are related
Main effects
A main effect in experimental design and statistics refers to the impact or influence of a single independent variable on a dependent variable, while holding all other independent variables constant. The main effect of variable A represents the average difference in the dependent variable C between levels A1 and A2, while disregarding the levels of B.
Main effects refer to the differences in mean scores between the levels of each independent variable across the values of the other independent variable (i.e., the marginal means).
Marginal mean
the marginal means of one variable are the means for that variable averaged across every level of the other variable.
Interaction
The effect of one independent variable depends on the level of another independent variable.
Interactions help identify situations where the effect of one factor (e.g., A) on the dependent variable (C) is different depending on the level of the other factor (B).
Researchers have 2 ways of conclusing wether they have found an interaction effect in a study:
- First way is by referring to the ANOVA summary table looking for the interaction effect and deterining its p value (<= 0.05)
- Plotting the interaction on a graph
When we plot an interaction graph how do we know we have an interaction?
The lines cross one another
effect size
meta analysis
a systematic review of previous studies in the same topic
One common way to increase the liklihood of finding statistical significance is to—– but—-
increase sample size
but
the effect size may be so small as to have minor practical value
Dr. Delgado’s computations showed that the eta squared value for one of the main effects in her study was quite large. This might mean that she has demonstrated:
a large effect size
Type 1 error
finding false positives or rejectinv a null hypoethesis that is true
What is the purpose of a post-hoc test after a statistically significant main effect finding for an independent variable with three levels?
to determine between which levels of the independent variable significance was found
Higher order factorial design
a type of factorial design in which the researcher manipulates 3 or more independent variable
When we have 3 independenr variables, we know we could have up to— statistically significgant main effects
3
How do you know you have a three way interaction?
When the 2 way interaction plot differ for each level of the third factor
- You know there is a three-way interaction because the plots in C1 and C2 are different.
Power
Power is the probability that you can detect an effect that is present. In other words, this is the probability that you will not make a type II error, or fail to reject a false null hypothesis. Power is usually stated as 1 - Β. Beta is usually set at 0.2, which results in the typical 80% power. The higher your power, the less likely you are to make a type II error.
In between subject designs, a simple way to gain power is to
increase sample size
When a study show statistically significant results yet tiny effect sizes, this means
that even though an independent variable might significantly affect a dependent bariable, its effect may explain very little of the varince in that dependent variable
80% power means
you will detect a real affect 80% of the time and miss a real effect 20% of the time
Mixed Designs+ ex (2)
includes elements of both between subject and repeated measures design
- ex: in a 2x2 mixed factorial design, one of the independent variables will be a repeated measures variable (exposed to both playing alone ot with eachother) and the other a between subject variable (what type of game)
Covariate designs
statistically control for a rival explanatory variable by running a statitical test to tease out the covariate.
- ex: when analyzing the dependency between a blood protein and a given diagnosis, age can be a covariate.
Multivariate Design
disordinal interaction:
In a disordinal interaction, the graph will show the lines depicting the one independent variable cross one another.
ordinal interaction:
In an ordinal interaction the graph will show that the lines depicting one independent variable do not cross, but are not parallel lines.
