1/65
Looks like no tags are added yet.
correlations ask
how do changes in the value of one variable related to the values of another
correlation coefficients
way of assigning one number to describe relationships (direction and strength), “r” -1 to +1
hypothesis and alternate hypothesis
H0: r=0, H1: r=/
how to decide if OV r is significant
compare to crticial
2 tailed test
alpha 0.05
DF is n-2 (sample size or # pairs)
why is n-2 DF for correlation
bc r is función of both x and y
reporting r
A correlation coefficient between x and y revealed a significant blah r(DF) = x, p<0.05. This finding suggests _______ (nature of relationship)
true or false: r is both an OV and ES
true
r as an ES vs as a significance test
OV of r is not influences by sample size, must compare it to CV (which is influenced by sample size)
CV or p value ___ as sample size ___
decreases, increases
coefficient of determination
how much 2 variables have in common, how much variation they share (%)
why cant we assume causation form a correlation
reverse causation, reciprocal causation, third variable
categorical variables
data that sorts data into categories ex: sex, party
continuous variables
data measured on a continuum, ex: weight, height, 1-7 likert scale
one way anova used for
single factor experiments, one independent variable, simple experimental design
factorial anova used for
multiple factor experiments, 2+ IV more than 2 groups,
number of numbers
number of factors, IV
value of the numbers
number of levels for each factor
number of conditions
multiply numbers in factorial notation , number of cells
main effects
overall effect of a single factor, ignoring all other factors
how to compute main effect
compute row and column means that combine data for all cells, doesn’t tell whether the main effect is stat sig
interactions
when the effect of one factor depends on the level of anotherm
main effect on a graph
one line higher than other, or one side higher (steep slope)
2 main effects on a graph
2 lines parallel in sabe direction
interactions on a graph
lines are not parallel
factorial anova is still an f test which means
F= variance between / variance within
between groups variance reflects differences due to, while within groups reflects differences due to
between: treatment effects and chance, within: chance
reporting f test
F (df for the source, df within) = f for the source, p
regression
stat tech for finding best line for a set of data, 2 continuous variables, minimize (residuals) distance between data points and the line
how does regression involve correlation
basically translates r into an equation that predicts the value of one variable (y, outcome) from the x (predictor)
y prime
predicted score
residual
distance between regression line and actual y score , y - y prime“error”
smaller our residuals ___ our predictions
better
how to find equation
solve for b first (slope)
interpretation of equation y=1x+ 5
for every one year increase we predict increase of 1 units of happiness, 5 is predicted value at age 0
unstandardized beta vs standardized beta
estimate , stand. estimate
unstandardized beta
hand calculated, predicted change or slope in natural units of predictor
standardized beta
tell predicted increase in SD units of predictor and outcome , compare across variables in different units ex: likert scale
ex: for every 1 sd increase in x, y increases _ sds
standard error of the slope
influenced by average size of all residuals
true or false p value changes whether you use standardized or unstandardized beta
false
reporting regression results
a regression equation revealed that x was sig or not sig predictor of y, b= x, p
regression vs correlation
correlation only 2 variables, regression: multiple predictors of a single outcome
equation for multiple regression
y prime = bX1 + bx2+ bx3 + a
NHST in multiple regression
each beta has its own hypothesis
h0: b(sleep)=0 OR H0 r²=0
each beta is an indicator of contribution of that variable
amount of increase in outcome assuming all other variables are constant
R² in regression
percentage of variance accounted for by all predictors together, effect size
parsimony
accept the simplest answer, variables that don’t provide improvement to prediction should be discarded
parametric tests
t tests and ANOVAs assume stuff of data such as
dependent variables: continuous
normal distribution
variances of groups are similar
Big enough sample
nonparametric tests
alternative tests when assumptions are untrue
ex: non-parametri
chi square
approximation of overall size of difference (between actual data and expected)
if no big difference chi square will be
close to 0
null hypothesis for chi square
null: P1= P2 = P3= P4, p is % of occurrences in each group
computing chi square
O- E,( O - E )squared, (O - E)²/E
DF for chi squared
r-1 rows in table or categories, horizontal
reporting chi square results
a chi square test revealed a(no) difference in something among the sample x² (df) =, p. More people (x%), , the second most (x%)
level of risk
0.05, alpha, 5%
when to use chi squared test
categorical variable measured as frequencies comparing observed counts against expected
when do we use anova
more than 2 groups compared on 1 DV
MANOVA
more than 2 groups and more than 1 DV
MANOVA answers
Does the factor have an omnibus effect on DV, does factor influence each DV individually
repeated measured ANOVA
ANOVA when factor is within subjects, follow same subjects over time
repeated measures is to factorial anova as independent t test is to
dependent t test
ANCOVa
controls for a variable we think is related to outcome/ DV
covariate
variable we control for
structural equation modeling
allows researcher to form stat model and test whether model fits data
meta anaylsis
way for researcher to combine data from any study, organize and summarize
factor analysis
technique used to determine whether measures form natural clusters or factors