research methods psy2023 Flashcards
what should the answers of pearsons correlation coefficient range between
r and -1 -> 1
how to work out sum of xy, sum of x and y squared
xy -> sum of the product of each x*y
sum x/y squared -> sum of each x squared individually
what is y = a + (b * x) used for
simple linear regression
best values go into these in order to predict later scores
relationship between a and b and b1 and b0
a may be b0 and b may be b1
the equation and process is still the same
may be displayed as y= b0 + b1x
what are residual values
the difference between an observed and expected value when predicting scores in linear regression
calculate observed-fitted
how to predict values in simple linear regression
put back into equation y=a + (b * x)
multiple correlation: how to work out
prop of variability in Y explained by x1 (r2)
p of v in y explained by x2 (r2)
p of v in y explained by x1 and x2(r2)
y and x1 = r01 squared
y and x2 = r02 squared
x1 ans x2 = multiple correlation = rsquared equation
what do you need to work out do do the equation in multiple correlation
r01, r02, r12 and their squares
when is r12 used in multiple correlation
only the equation, it is not proportion of variability in y explained by x1 and x2 together
how to work out a, b, sa, and sb in sobel test
a = effect of x on m (dv=m)
b = effect on m on y (dv = y)
sa and sb of the standard errors
what do you need to know in sobel before doing the equation
a, b, a squared, b squared
sa, sab, sa squared and sb squared
make sure negatives are removed when squaring minus numbers
how to use lookup table in correlation coefficient
degrees of freedom/numbers on left is N
how to assess bias
risk -> avoiding systematic error
control
selection, performance, detection bias and reporting bias
how do you use literature to support a mediation model
need 3 relationships
predictor -> outcome
predictor -> mediator
mediatory -> outcome
types of mediation
full: entirety of association between y and y is explained by m
partial: some of association between x and y explained by m
types of probability based sampling
random
stratified random -> stratify into homogenous groups then simple random
cluster -> natural sampling unit is a group (secondary schools)
systematic -> every nth person from a sampling frame (every 10th to enter website)
non probability sampling types
convenience
quota sampling - first 20
snowball - referrals
judgement - based on judgement, sample from representative subgroup
errors of coverage - sampling
populations of inferences -> who the conclusions are for
target pop - pop of inference - disregarded groups
frame pop - portion of target pop that may be studied
coverage error is the difference between frame populations and population of inference
how to reduce coverage error
obtain as complete as sample frame as possible, or be frameless
post stratifying - weight based on population of inference
assumptions for regression models
relationships are linear
errors assumed to be independent (random scatter), normally distributed and with common varience (scatter similar to all points with no obvious patterns)
how to test significance of linear regression
test null hypothesis
use one sample t test
if a straight line is suitable to model the data
need slope of regression line to be 0
what is the meaning of r squared
the proportion of variability in percentage that measures the proportion of the variance in a dependent variable that is explained by the independent variable(s) in a regression model
what the the normality tests
kolmogrov-smirnov and shapiro-wilk
need p values to be non-significant to show that there is no evidence for them not to be normally distributed
need normal distribution
what are we interested in in mediation
a - whether a direct effect of x on y exists
b - whether indirect effect of x on y via m exits
c - whether an additional effect of x on y exists having allowed for m
what is the casual steps method/baron and kenny’s method
- find effect (significance) of X in a simple linear regression on Y
- find the effect (significance) of X in a simple linear regression on M
- find the effect (significance) of m in a multiple regression on y with x
- find the effect (significance) of x in a multiple regression on Y with M
meaning of results from causal steps method
1 tells us about a
2+3 tells us about b
4 tells us about c
if x is sig in 1 but not 4 it means m mediates the relationship between x and y
what is the simple regression equation in formula and spss terms (step one)
y = a + (b * x)
dependent variable = unstandardised coefficient b constant + USB variable x * variable x
if it is significant x predicts y
what is the regression equation for step 2 in spss output
dependent variable (m) = UCB constant + x variable UCB * grade variable
multiple regression equation in spss
y = ucb constant + ucb variable x * x + ucb variable m * m
issues with causal steps/baron and kennys method
5% cut off not ideal to measure mediation relationships with
alternative to causal steps
using indirect effect
ie = a * b
where a = regression coefficient using x to predict m
b = regression coefficient for m when using x and m to predict y
use confidence intervals to assess significance
what is a confidence interval
range of values that the true values would lie in 95% of the time
common limitations
design -> causation, limited sample application
data collection -> bias
analysis -> power, clear reporting
results -> internal or external validity
what is moderation for
to show how the association between predictor and outcome can vary depending on another variable
association between x and y depends on m
difference between mediation and moderation
med -> x affects y, and x affects m which affects y
mod- association between x and y differences depending on m
how is moderation visualised on a graph
when groups diverge, converger or cross
not when parallel
how do effect size, varience and sample size affect significant
es- larger the diff between two groups, more likely the samples diff
variable - larger varience, less likely to be sig
ss - smaller samples more easily weighted
how to centre variables moderation
variable - mean of variable
how to write moderation equation
same as linear with y = a + (x * variable) + (x * variable) + ( x * variable1 * variable 2)
how to know if there is a mediation model
if theres a stronger association between x and y when m is involved = mediation
when just x and y is stronger - no mediation and direct effect
how to write up mediation model
state association
write equation
state significant
state what it means
state r squared and the % of variation that the variable counts for
copy for each and then and the end sum up if there’s a model or not
how to write up moderation
regression equation
what variables contribute to to explaining y and their significance
say whether the interaction of x and m is significant, and therefore is there a moderation model
what to put into intuitive explanations
what results for x and m alone may mean for y
how to read spss results for peason’s correlation
look at sig. 2 tailed value
how to report pearson’s correlation
r squared
n value
significance
evidence to reject null? state null
