Term
Chi Square Terms goodness of fit
2 kinds |
|
Definition
for nominal data
2 assumptions: 1) independence of observation, 2) mutually exclusive and exhaustive categories
tells whether observed frequencies match expected
cannot use with time series/repeated measures designs
single-sample (one variable) and multiple-sample (multiple variable) versions - number of variables INCLUDES IV and DV |
|
|
Term
|
Definition
Tests whether means of several groups are different. extends t-test to more than 2 groups.
Uses the F-statistic. Significant when > 1
F-ratio = MSB/MSW = (Treatment + Error)/Error
use post-hoc tests to identify where the difference lies (X, Y, XY intxn)
- Scheffe, Tukey, Fischers |
|
|
Term
|
Definition
| compares a single sample to a known population mean to see if sig diff |
|
|
Term
|
Definition
when have more than one continuous IV w/ on continuous DV
also called Factorial ANOVA
"it depends" - interaction effects when significant |
|
|
Term
|
Definition
| when have more than one DV (may or may not have more than one IV) |
|
|
Term
| correlation measure for curvilinear relationships |
|
Definition
|
|
Term
| homoscedasticity vs. heteroscedasticity |
|
Definition
homo - equal scatter around regression line
hetero - unequal scatter around regression line (one portion of group might be further away from line while another is clustered around it closely) |
|
|
Term
|
Definition
| overlap in predictive value of different IVs - you dont want this |
|
|
Term
types of data:
nominal
ordinal
interval
ratio |
|
Definition
nominal - categories/names (nonparametric)
ordinal - rank order (nonparametric)
interval - equal distance between data points, e.g., temperature. No absolute zero (parametric)
ratio - equal distance between data points, can be add, sub, mult, divide. has an absolute zero(parametric). |
|
|
Term
| central limit theorem states |
|
Definition
| as sample size increase, distribution of scores become a normal curve |
|
|
Term
| Standard Error of the Mean |
|
Definition
| Standard deviation of the population mean --> tells error in your prediction of population mean from your sample. "the extent to which any one sample randomly drawn from the population can be expected to vary from the population mean...it is a measure of variability due to random error" calculated by dividing SD of population by square root of sample size (N) |
|
|
Term
|
Definition
increasing alpha
increasing sample size
decrease extraneous variables (control for or covary out)
increase ES (maximize intensity of IV)
1-tailed test
parametric test |
|
|
Term
What are the nonparametric siblings of these tests?
t-test correlated groups
t-test independent sample
ANOVA |
|
Definition
corr t-test = Wilcoxson Matched pairs
ind samp t-test = Mann Whitney U test
ANOVA - Kruskall-Wallis |
|
|
Term
| common assumptions of parametric and nonparametric, differences, and shortcomings |
|
Definition
Common:
random selection of sample
independence of observations
Difference:
Parametric - normal distribution of data and homoscedasticity
Nonparametric tests are not as powerful |
|
|
Term
| pearson product coefficient |
|
Definition
correlation coeff (r)
for interval or ratio data, tells straight up correlation |
|
|
Term
| Spearman Rank-Order (rho) |
|
Definition
| tells correlation of for ordinal or rank data |
|
|
Term
|
Definition
| tells correlation of true dichotomy variable (sex) |
|
|
Term
|
Definition
| tells correlation of artificial dichotomy (high earners and low earners) |
|
|
Term
|
Definition
| tells correlation of NOMINAL data (e.g., DSM diagnosis) |
|
|
Term
|
Definition
uses multiple continuous or discrete predictors to predict status on a continuous criterion variable.
gives a multiple correlation coefficient (R)
Types:
simultaneous - put all variables in at once
stepwise (add or subtract one variable at a time and see how R changes, stop when R-squared is high enough) |
|
|
Term
|
Definition
| using a Mult Reg equation on a new sample. likely to get shrinkage of correlation coeff bc the same random error is not present in the second sample as in the first one |
|
|
Term
| discriminant function analysis |
|
Definition
| trying to predict an individual's status on a certain category/nominal criterion using 2 or more continuous predictors |
|
|
Term
|
Definition
| extension of mult reg with 2 or more continuous predictors to predict status on 2 or more continuous criteria |
|
|
Term
| reject when true is not type II! |
|
Definition
type 1 - reject a false null hypothesis (say there is an effect when there is not)
- determined by alpha
type II - retain a false null hypothesis (say there isnt an effect and there is)
- beta, cannot be known or controlled for. inverse relationship with alpha |
|
|
Term
|
Definition
• Multiple Analysis of variance (MANOVA)– used to control for Type I error
• Analysis of Covariance (ANCOVA)– using your statistical analysis to control for a compound variable
o Take out effect of a variable that is going to impact the study. (Ex.30 students in reading class, can take out a subject whose IQ will impact the testing)
• Randomized Blocked ANOVA (Block Design) – using a variable by classifying in your groups and in your study
o Three groups classified by reading aptitude (blocking part), want to improve reading aptitude (variable)
• Split-plot ANOVA – split design, works with between and within study.
o Mixed-design – within group and between group |
|
|
