Research Methods Modules Flashcards
difference
is one group of people different to another in some way?
association
is one construct related to another?
prediction
does one construct influence another?
goal of psychological research
to make inferences about a population (inferring that what is typical for sample is typical for population)
population
everyone of interest to a research question
distributions of data can be described according to their
central tendency (eg. Mean) and variability (eg. Standard deviation)
the normal distribution
majority of observations in the middle, observations reduce in frequency towards the tails, the distribution is symmetrical
in a normal distribution, most observations are closed to m; these scores occur more
frequently; typical
the 2s Rule of Thumb
in a distribution with a normal shape, 95% of scores fall within approximately 2 standard deviations from the mean. These scores are typical
typical scores
are expected and occur frequently
extreme scores
are not expected and occur infrequently
distribution of sample means
made up of the sample means from all of the random samples of a certain size (n) that could possibly be obtained from a population
Central Limit Theorem tell us
the precise characteristics of a distribution of sample means for samples of any size (n). The distribution of sample means has equal mean to the population mean, for large sample sizes, the distribution of sample means will be normal, details of standard error, as sample size increases, standard error decreases and estimation of population mean becomes more precise
standard error
standard deviation of the distribution of sample means
when sample is large enough, it provides
a reliable estimate of the population mean
z-test standard error formula
standard error = standard deviation of population/(number of people in sample)^1/2
we can use 2s Rule of Thumb to test if our SAMPLE MEAN
is typical or extreme
hypothesis
a statement that predicts that something is going to happen
experimental hypothesis/alternative hypothesis
a statement that predicts an effect (one of difference or association)
null hypothesis
predicts that nothing is happening; a hypothesis of no effect (no difference, no association)
only one of null/experimental hypothesis can be
supported by research data at any one time
null hypothesis statistical notation
H0
experimental hypothesis statistical notation
H1
null hypothesis significance testing
propose a null hypothesis that a population parameter (mean) has a particular value. Proceed assuming the null hypothesis is true. Determine the probability of the sample mean occurring if the null hypothesis is true. If the probability of the sample mean occurring is small, reject the null hypothesis. If the probability is large, do not reject the null hypothesis
if the probability of the sample mean occurring is small,
reject the null hypothesis. Evidence for a difference. Extreme sample mean
if the probability of the sample mean occurring is large,
do not reject the null hypothesis. No evidence for a difference. Typical sample mean
determine the probability of the sample mean occurring if the null hypothesis is true. In other words,
what is the likelihood of our sample mean occurring if the mean of the population really is the value we predicted in the null hypothesis?
how to determine the probability of the sample mean occurring if the null hypothesis is true
involves a statistical test based on a normal distribution of sample means with the mean we predicted in our null hypothesis. Calculating critical limits to determine if our sample mean is typical or extreme
the 5% Alpha Level
defines which sample means in a distribution of sample means are expected or typical, and which are unlikely or extreme, if the null hypothesis is true
when the comparison distribution is perfectly normal, the critical limits set by the 5% Alpha Level are precisely
+/- 1.96 standard errors from the mean. 95% of the scores are inside these limits
if our sample mean is inside the limits set by the 5% Alpha Level, the probability is
greater than 5%, and therefore high (do not reject the null hypothesis)
if our sample mean is outside the limits set by the 5% Alpha Level, the probability is
less than 5%, and therefore low (reject the null hypothesis)
single sample z-test
how we determine the probability of our sample mean occurring after setting an Alpha Level of 5%
z-score
how many standard errors our sample mean is away from the null hypothesis
single sample z-test formula
z-score for sample mean = (sample mean - population mean)/(z-test standard error)
in single sample z-tests, the population standard
deviation is known
once z-score has been calculated, check whether it is more extreme than
+/- 1.96
if z-score is more extreme than +/- 1.96, the probability of sample mean occurring assuming the null hypothesis is true is
less than the Alpha Level of 5%, so probability is low and null hypothesis is rejected
steps for determining whether sample mean provides evidence to support null hypothesis or not
set Alpha Level (5%) then calculate z-score
we can’t use a single sample z-test and the normal distribution when the
population standard deviation isn’t known
we use the t-test and ‘t-distribution’ when
the population standard deviation isn’t known
in single sample t-tests, we use the sample standard deviation as an
estimate of the population standard deviation
t-test standard error formula
standard error = standard deviation of sample/(number of people in sample)^1/2
t-test formula
t-score for sample mean = (sample mean - population mean)/(t-test standard error)
almost all aspects of the process are the same when conducting a
single sample z-test or t-test
in the t-distribution, critical limits corresponding to Alpha Level of 5% will not be fixed at
+/- 1.96 as in z-test
t-distributions require that we consider
sample size and degrees of freedom (df)
degrees of freedom
one less than our sample size for single sample t-test (n-1)
critical limit in single sample t-test varies
along with df
to check if t-score is more extreme than critical limit taking df into account, use
SPSS or look up in back of textbook
Alpha Level of 5% still applies in t-tests, it’s just that we can’t automatically assume
critical limits of +/- 1.96
test value in SPSS value is the
null hypothesis value
sig in SPSS output is the
probability of our sample mean occurring (in %)
correlational research design
examines relationship between two variables
in correlational research, each participant provides how many pieces of data?
2
in correlational research, a ______ observation is made
simple
in correlational research, there is no
control or manipulation
correlational research examines
ASSOCIATION
positive linear association
as scores on x increase, scores on y also increase. High scores on x are related to high scores on y
negative linear association
as scores on x increase, scores on y decrease. High scores on x are related to low scores on y
linear trends are barely or not observable if
weak or no association exists
correlation does not tell us about
direction of effect/causation
in correlation, there is no
IV or DV
in correlational research, it is possible that another
variable explains an observed relationship
Pearson’s Correlation Coefficient (r)
measures the strength of the linear relationship between x and y
the value of r lies between
+1 and -1
the size of r specifies how close the data is to
a straight line
the sign of r (+/-) specifies the
direction of the association
.10 r value
weak (small effect)
.30 r value
moderate (medium effect)
.50 r value
strong (large effect)
r values close to 1, whether negative or positive, tell us that the correlation is
quite close to being a straight line
r values of zero mean
no correlation, so scores closer to zero are weaker
Pearson’s r procedure in JASP tells us
the strength and direction of the correlation, and if we can infer that an association observed in a sample (r) is also present in the population (‘rho’)
if r is large enough (so that it is extreme in a distribution of sample correlation coefficients), we can
infer an association between two variables in a population and reject the null hypothesis
rho (fancy p)
correlation in the population
null hypothesis for Pearson’s (r) analysis
rho = 0
experimental hypothesis for Pearson’s (r) analysis
rho is not zero
for Pearson’s (r) analysis in JASP, if statistical significance (p) is less than 0.05, we can
infer correlation in population
independent groups research design
participants are assigned to, or come from, two or more different groups
research question for independent groups research design
is there a difference between the groups?
independent samples t-test
is there a significant difference between the means of the two groups?
null hypothesis for independent samples t-test
no effect, no difference between means of two groups. The population mean of each group would be equivalent
experimental hypothesis for independent samples t-test
of effect. The population mean of each group would be different (one bigger than the other)
the experimental hypothesis in an independent samples t-test is actually predicting
that the two groups come from different populations
in an independent samples t-test, if the p-value is <0.05,
reject the null hypothesis, a statistically significant difference has been found
Cohen’s d
effect size
little overlap between groups (independent samples t-test)
evidence of different populations
lots of overlap between groups (independent samples t-test)
evidence of same population
independent samples t-tests assess if
the difference between two sample means is different to zero OR if one sample mean is < or > than another
the JASP output for independent samples t-test gives
descriptives, t-score, p-value, effect size, and more
the JASP output for independent samples/repeated measures t-tests gives
descriptives, t-score, p-value, effect size, and more
control groups
do not receive the experimental treatment
types of t-tests
single sample t-test, independent samples t-test, repeated measures/paired samples/related samples t-test
repeated measures research design
each participant is measured on two or more different occasions
research question for repeated measures research design
is there a change across time?
in a repeated measures research design, we use the same measurement/test
at times 1 and 2 to ask ourselves if the mean scores of the two samples of data are statistically significantly/meaningfully different
repeated measures t-tests are also known as
paired samples or related samples t-tests
repeated measures t-tests
is there a significant difference between the means at time 1 (before) and time 2 (after)? Whether the difference between the T1 and T2 means (on the same measure) is different to zero OR if one sample mean is < or > than the other
null hypothesis for repeated measures t-test
mean of population at T1 is same at T2; mean of population at T2 subtracted from T1 equals zero
experimental hypothesis for repeated measures t-test
mean of population at T1 minus mean of population at T2 will not equal zero
for repeated measures t-test JASP data, if the middle score for the second boxplot is higher/lower,
the median has increased
in a repeated measures t-test, if median difference in difference scores is
higher than zero, the mean of different scores is different to zero; the treatment works
when we find a significant difference with regard to the construct in a repeated measures t-test,
there is evidence of different populations (less overlap)
when we don’t find a significant difference with regard to the construct in a repeated measures t-test,
there is not evidence of different populations (more overlap)
number of sample groups in single sample research design
one
number of measurements taken from each group in single sample research design
one
single sample research design measures/compares what?
sample to population mean
repeated measures research design measures/compares what?
difference between T1 and T2
number of sample groups in repeated measures research design
one
number of measurements taken from each group in repeated measures research design
two
independent groups research design measures/compares what?
group 1 and group 2
research question example for independent groups
does G1 differ from G2?
research question example for repeated measures
does T1 differ from T2?
research question example for single sample research design
does sample differ from population mean?
number of sample groups in independent groups research design
two
number of measurements taken from each group in independent groups research design
one
correlational research design measures/compares what?
extent to which two variables co-occur
research question example for correlational research design
is V1 associated with V2?
number of sample groups for correlational research design
one
number of measurements taken from each group correlational research design
two
reliability
does the measurement yield consistent, dependable, and error-free information
validity
does the measurement assess what it is intended to assess and is it useful
internal consistency (part of reliability)
do the components of the test all cohere? All test items should correlate with each other
inter-rater reliability
does the test give the same information about the person when different people administer it?
re-test reliability
does the test yield similar scores when it is administered to the same person on different occasions?
high reliability =
high consistency = low measurement error
does the test measure what it is intended to measure?
content, convergent, and discriminant validity
does the test provide practically useful information?
predictive validity
if reliability is low,
validity cannot be high
unreliability exists when
there is inconsistency in what the test measures
invalidity exists when
the test does not measure what it should