Research Design Flashcards
Predictive value
the degree to which an independent variable can predict a dependent variable
Cohort effects
the effects that might result when a group is born and raised in a particular time period.
Demand characteristic
when subjects act in ways they think the experimenter wants or expects
Experimenter bias (another name)
Rosenthal effect
Hawthorne effect
when subjects alter their behavior because they’re being observed
Selective attrition
when the subjects that drop out of an experiment are different from those that remain. Remaining sample is no longer random
Social desirability
subjects do and say what they think puts them in a positive light.
Frequency distributions
explain how the data in a study looked. Might show how often different variables appeared.
Nominal variables
male/female; Republican/democrat
Ordinal variables
implies order; not necessarily equally spaced (marathon finishers)
Interval variables
show order and space because equal spaces lie between the variables, but do not include a real zero (temperature)
Ratio variables
order, equal intervals, and a real zero (age)
Frequency polygon
plotted points connected by lines; used to plot continuous variables (categories without clear boundaries)
Histogram
vertical bars; sides touch; discrete variables with clear boundaries and interval variables with some order
Standard deviation
(1) subtract mean from each; (2) square each; (3) add; (4) divide by original number of scores; (5) take square root
If standard deviation is large – scores highly dispersed
If small, scores very close together
Unimodal
only one hump
Z-score
how many standard deviations a score is from the mean; -3 to +3 for normal distributions
T-score
transformation of a z-score in which the mean is 50 and the standard deviation is 10; T = 10(z) + 50
Standard normal distribution
same as normal distribution but standardized so that the mean for every distribution is zero and the standard deviation is one
platykuric distributions
flat top/same sides
Pearson r correlational coefficient
- 1 to +1
Spearman r correlational coefficient
used only when data is in the form of ranks – used to determine the line that describes a linear relationship
Statistical regression
allows you to not only identify a relationship between two variables but also make predictions about one variable based on another variable
Parameters
refer to numbers that describe populations
statistically significant
describes a real pattern
Alpha level
significance level; < .05 or < .01 (5/100 or 1/100)
Type 1 error
incorrectly reject the null hypothesis (thought findings were significant but they weren’t)
Type 2 error
wrongly accept the null hypothesis (say no relationship, but there is)
T-test
compare the means of two different groups to see if the two groups are truly different; used to analyze means on continuous data; particularly useful when samples are small; cannot test for differences between two groups
Chi-square tests
used when n-cases in a sample are classified into categories (cells); tells us whether the groups are statistically significant in size; analyze categorical or discrete data; can be used on small samples; can assess goodness of fit of distribution or whether the pattern is what would be expected
ANOVA (Analysis of variance):
analyzes differences among means of continuous variables; can analyze difference among more than two groups
One-way ANOVA
tests whether the means on one outcome or dependent variable are significantly different across groups
Two-way ANOVA
tests effects of 2 independent variables or treatment conditions at once
Factorial analysis of your variance
Used when an experiment involves more than one independent variable; can isolate main effect; can identify interaction effects
Analysis of Covariance (ANCOVA)
• Tests whether two groups co-vary; can adjust for preexisting differences between groups
Linear regression
Use correlational coefficients as a predictive measure; based on a line fit with the least-squares method
standardized
tried out on huge groups of people to create norms
Criterion-referenced tests
measure mastery in one particular area (final exam)
Domain-referenced tests
attempt to measure less defined properties (like intelligence); need to be checked for reliability and validity
Reliability
how stable the measure is
Split-half reliability
comparing an individual’s performance on two-halves of the same measure (odd vs. even); reveals internal consistency
Item-analysis
how a large group responded to each item on a measure
Concurrent validity
whether scores on a new measure positively correlate with other measures known to test the same construct
• Cross-validation
Construct validity
whether the test really taps the abstract concept being measured
Content validity
whether the content of the test covers a good sample of the construct (not just a part of it)
Face validity
test whether the items simply look like they measure the construct
Donald Campbell and Donald Fiske
multi-trait-multi-method technique to determine the validity of tests
