Research Methods & Stats Flashcards
Item response theory/latent trait theory
used to calculate to what extent a specific item on a test correlates with an underlying construct. Can be used to equate scores from non-parallel measures. Used to develop individually tailored adaptive tests.
classical test theory
individual’s test score is the sum of true score variability and error score variability
empirical criterion keying
Used in MMPI development, items chosen based on their ability to discriminate group membership.
ipsative measures
yield information only on the individual and not on how that individual compares to others
Partial correlation
correlation between two variables when the effects of a third variable have been partialed out (removed) from both variables
Semi-partial correlation
correlation between two variables when the effects of a third variable have been partialed out (removed) from one of the variables
coefficient of determination
proportion of variance shared by two variables and is the square of the correlation efficient
Interval recording
a type of behavioral sampling used when there is no clear beginning, middle, and end of a behavior. Breaking up a period of time into smaller parts and recoding whether or not behavior took place during each interval. A type of time sampling
Standard error of measurement
average amount of error in predictor variable
Cluster sampling
identifying naturally occurring groups and then randomly selecting certain groups and sampling all within group or random sampling of group
When is pooled error term used?
there is homogeneity of variance, i.e., homoscedasticity
When is a separate error term used
there is heteroscedasticity, that is variance is not equal
1-way ANOVA variables
1 IV, 1 continuous DV
MANOVA variables
multiple DVs
Chi-square variables
nominal DV
2-way ANOVA
variables
2 IVs, 1 DV
3-way ANOVA variables
3 IVs, 1 IV
Factorial ANOVA
More than one IV, 1 DV
Correlations are weaker when
range is restricted
Assumptions for chi test
independence of observation (e.g., cannot be used when repeated measures are made)
When to use MANOVA?
Multiple DVs
formula for standard error of the mean
= population standard deviation divided by the square root of sample size
analysis of covariance (ANCOVA)
combines standard analysis of variance (ANOVA) with the technique or partial correlation. Allows effects of an extraneous/moderator variable to be statistically removed (partialed out).
exclusively used when partialing out the effects of a confound variable from research looking for differences between groups, not relationships among variables
When is a t-test appropriate?
when there is only one IV and only two groups are being compared
assumptions of factorial ANOVA
more than one IV
all IVs are between groups
data are independent
define homoscedasticity
there is equal variance among groups
Define partial correlation
a statistical procedure that investigates the relationship between two variables while controlling (partialing out) the relationship of a 3rd variable
Percentage that falls within 1 SD of the mean (normal curve)
68%
Percentage that falls within 2 SD of the mean (normal curve)
95%
Percentage that falls within 3 SD of the mean (normal curve)
99.7%
What are z-scores?
Measure the distance for raw data from the mean in terms of standard deviations
formula for z-scores
Z= (x-Mean) divided by SD
What stanine score (of raw scores) is roughly equivalent to the mean?
5
Type I error
rejecting the null hypothesis when it’s actually true (concluding results are significant when in reality they came about by chance)
alpha
Type II error
Not rejecting the null hypothesis when it’s actually false; failing to conclude there was an effect when there actually was (perhaps due to lack of power)
Beta
What level of statistical power is considered sufficient
80% or higher
How is statistical power determined?
1-beta
What are beta weights?
standardized regression coefficients
What do beta weights signifiy?
the strength of the relationship between the predictor and criterion in standardized regression
