Chapter 11: Complex Experimental Designs & Inferential Stats* Flashcards
Curvilinear relationship
requires at least 3 levels of the independent variable
Complex (factorial) design
research with 2 or more independent variable (i.e. factor) and with each IV also having more than one level; Number of levels of first IV x Number of levels of second IV
Factor
any outcome that you expect to be related to some outcome variable e.g. IV that affects DV
What 2 distinct kinds of information do factorial designs yield?
main effect and interaction
Main effect
the direct effect of an IV on a DV, ignoring any interaction with other IVs
Interaction
When the effect of one IV on the DV depends on the level of another IV; parallel=no interaction; not parallel=interaction
Marginal mean
average score of all participants in one condition of one independent variable, collapsing across the levels of the other IV
Moderator variable
third variable that influences the relationship between an IV and DV; effect is revealed as interactions
Simple main effect
effect of one IV on the DV at one particular level of another IV; mean difference at each level of one IV
IV x PV design
factorial design that includes both an experimental IV and a non-experimental participant/person variable; allows researchers to investigate how different types of people respond to the same manipulated variable
Mixed factorial design
a factorial experimental design that includes both between-subjects and within-subjects variables (IV x PV)
Inferential statistics
statistics that estimate whether the results observed based on sample data are generalizable to the population from which that sample was drawn
Statistically significant
observing that an outcome has a low probability of occurrence (p-value < .05), assuming that H0 is correct; difference between groups reflects a real difference in population
Significance or alpha level
0.05; how willing you are to be wrong if you conclude there is an effect in the population; threshold probability which a test statistic is deemed to be unlikely to have come from sampling distribution
Steps in null hypothesis significance testing (NHST)
(1) Formulate the null hypothesis and assume H0 (2) Collect data (3) Calculate p-value of getting such data or even more extreme data (4) Decide whether to reject or retain H0
Null “dull” hypothesis
baseline conservative assumption that IV does not affect DV; No relationship between 2 variables
Research hypothesis
IV does affect DV; there is a relationship between 2 variables (correlational) or there is a difference in means between experimental groups
What are the 2 characteristics of null and research hypothesis?
mutually exclusive (cannot have both be true at the same time) and exhaustive (must account for all possible outcomes)
High p-value
> 0.05; fairly likely that outcome was gotten from chance alone; retain H0
Low p-value
< 0.05; not likely that outcome was gotten from chance alone; reject H0
Sampling distribution
frequency distribution of values obtained if a study was repeated an infinite number of times, using the exact same parameters; done to evaluate likelihood of given result based on chance alone
t-test
NHST statistic used to compare two means; t= group difference (difference between obtained means)/within-group variability
F-test
NHST test for whether 2 or more means differ in the population
True state of affairs
truth about the null hypothesis if you could sample the whole population
Type I error
false positive; falsely reject the null hypothesis when it is actually true; error rate= alpha
Type II error
false negative; retaining the null hypothesis when it is actually false; error rate= beta
Power
probability of correctly rejecting the null hypothesis using a particular statistical test; directly related to type II error
What 3 factors do power and type II error rate depend on?
sample size, magnitude of effect (effect size), and alpha level
Sample size and power
greater sample size, greater power, less error in data to detect effect
Effect size and power
larger the difference is in population, easier to detect, thus greater power
Alpha level and power
larger alpha level, easier it is to find data consistent with research hypothesis (to reject null), thus greater power
P-hacking
a set of ethically questionable practices researcher uses to get statistically significant results
Obtained value
in a significance test, a statistic that captures the effect observed in your study; t=difference between means of conditions/(square root of sum of variances/sample sizes)
How do you reduce noise to get larger t-values?
consider poorly worded questions, effect of uncontrolled variables, small sample sizes, between vs. within subjects designs
Universe of possible values of t
sampling distribution of t or t distribution
Relationship between p-values and t-obt; alpha levels and critical values
p-value is the amount of area in the t-distribution that corresponds to a particular value of t-obt; alpha levels correspond to critical values
When do your reject or retain H0 in terms of t-obt and t-crit?
Reject H0: t-obt > t-crit; Retain H0: t-obt < t-crit
