purpose of method section
#NAME?
order of method section
by section:
1. participants - subjects if animals, participants
if humans
2. materials OR measures -
measures if you only used questionnaires, materials if you used
anything else- apparatus section only if you
used special equipment
3. design and procedure
method section: participants
1st section of method section
- describes your sample: give N (sample size), relevant
demographics, are range, average age and SD (include units ex: years,
months, weeks old), # men/women, special circumstances
- inclusionary/exclusionary criteria (what did/did not allow
someone to participate in the study)
- report things like: twins, marital status, ethnicity/race,
religion, socioeconomic status, education level if they are relevant
- if you used humans, you must describe them as participants - use
"subjects" only if you used animals
method section: materials or measures
2nd section of method section
- use "measures" if you only used questionnaires, use
"materials" if you used anything else (ex: computers, blood
draw, etc)
- for materials: describe the materials or
equipment used to collect your data, and include any relevant details
- computer, website, instruments used,
- for measures: describe the type of questionnaire
- T/F, fill in the blank, any scales you used, anchor labels (the
extremes on a scale, ex: 1 = strongly agree, 5 = strongly
disagree)- describe sample items if used and their
reliability/validity, if you create your own: state which current
measure you are comparing it to and where the items came from
- describe apparatus if you used specialized
equipment - describe any equipment and software used in detail if
readers are not familiar with it
method section: procedure
3rd section of method section
- start with a sentence or two on the design of
the experiment: independent/dependent variables, manipulated
variables/correlational or quasi-experimental design
- a detailed description of exactly what you did during your
experiment: how you recruited and assigned participants to groups,
what you told participants, type of environment used, where you
collected data, consent, debriefing, time allotted, actions
participants performed, rewards for participation, etc
- detailed enough so that someone can replicate your experiment,
but do not include irrelevant detail (ex: type of chairs participants
sat in)
other names for dependent t test
#NAME?
when is a dependent t test run
correlated/matched groups design - when you have two
groups that are matched (ex: identical twin studies)
repeated measures design - each participant is
measured twice on a variable
within subjects design - each participant is exposed
to two experimental conditions (ex: 2 levels of IV or
control/experimental levels)
paired samples (matched/correlated samples/pairs) -
two sets of scores that are related in some way (ex: 2 groups matched
on all characteristics)
basics of dependent t test
- you have 2 scores per person (or 2 matched people each with one
score), and you find the difference in the 2 scores
- the difference score is the score of interest, so in SPSS you make
a 3rd column with the difference in the 2 scores
null: mu1 - mu2 = 0alternative:
mu1 - mu2 does not = 0
- similar to t test for single samples (independent t test) except
you use the difference scores instead of raw scores
value of mu in dependent t test
mu is the mean of the population of difference scores, usually will
be 0 (unless it's diet pill or something where the population mean
shouldn't be 0)
it is usually 0 because you want the null hypothesis to be 0 so that
if you get any score higher than that you can compare the 2
degrees of freedom for dependent (paired samples) t test
df = N - 1
where N = # of paired scores
this is because you cannot run a dependent samples t test with
unequal N, all of the scores need something to be compared to so you
can find the difference in scores
the only values that are free to vary are 1 set of scores
what does an independent t test determine
it determines if there is a statistically significant difference
between the means of two unrelated groups - ex: female vs male
unrelated groups means that a member of group 1 cannot also be a
member of group 2
null and alternative hypothesis for independent t test
null: u1 = u2
for directional hypothesis (one mean will be
higher/lower than the other)alternative: u1 >
u2 OR u1 < u2
for nondirectional hypothesis (there will be a
difference between the two means)alternative: u1 ? u2
designing/running independent t test
- the means of 2 unrelated groups are compared to see if there is a
difference between them, or if one is higher than the other
- the mean of scores for members in group 1 is compared to the mean
of scores for members in group 2
characteristics of samples in an independent t test
value of mu and why is equals what it does
the 2 samples are unrelated - a member of one group cannot also be a
member of the other group
samples are taken from the population, which allows us to estimate
the population parameters (mu) because the values should be the same
(if they come from the same place, their averages should be the same)
mu is the mean of the population, which we are estimating from our
samples. we expect mu (for the difference between the two samples) to
be 0
standard error
tells us on average, how much error there is when we use sample means
to reference the population means
aka: when we sample scores from the population and use those scores
to estimate the real value of the population, there will be some error
the larger the sample sizes the smaller the standard error because
larger samples encompass more of the actual population so they are
fairly close to the actual population score
degrees of freedom for independent t test
dftotal = df1 + df2
dfT = N - 2
df1 = n1 - 1df2 =
n2 - 2
- this is different from the df for dependent samples because the
sample sizes in a dependent test must be = so df is just the # of
paired samples - 1 or N -1- in an independent t test the samples
do not have to be equal because the samples are unrelated so the
values that are free to vary are one per sample, so N - 2 for 2 samples
assumptions for independent t test
1. the sampling distribution is normally distributed (the populations
from which the samples were taken should be normally distributed OR N>30)
2. there is homogeneity of variance (the variances of the 2
populations being sampled are equal)
weighted or pooled variance estimate
pooled estimate - a way to estimate the variance of several different
populations even if the means of each population might be different
the pooled estimate gives us a more precise estimate of the
population variance than the individual sample variances - this leads
to higher power
weighted average - takes into consideration if the population sizes
are not the same and weights the estimates accordingly
power of independent vs dependent t test
dependent t tests have a higher power than independent t tests
because you are controlling for unexplained variability (error),
because you are basically comparing you to you, or in a matched groups
design, the groups are very similar so you control for outside factors
what does ANOVA stand for
analysis of variance
when do we use ANOVA
when you have 2 or more groups or 2 or more levels of the IV
what does ANOVA allow us to test for
allows us to test for a difference between the groups, when there are
2 or more groups, and tells you if there is an overall difference
between them or not
(does not tell you exactly which groups are different from each
other, thats what post hoc tests are for)
type 1 error (alpha)
- rejecting the null hypothesis when it is actually true (false
positive)- finding significance when there really is none
when alpha gets smaller (harder to get a significant result but
stronger results if you get significance) the probability of a type 2
error increases
type 2 error (beta)
- accepting the null hypothesis when it is actually false (false
negative)- not finding significance when the result actually is significant
power + type 2 error = 1
power
- probability of rejecting the null hypothesis when it actually is
false (true negative)- likelihood of a true significant result
bigger effect size = more power
power = 1 - beta (type 2 error)
boferroni correction
when you run multiple t tests the probability of a type 1 error
increases with each one so you use this correction to control for your
alpha level
alpha / c
null and alternative hypothesis for ANOVA
null: u1 = u2 = u3 - all groups are equal
alternative: u1 ? u2 ? u3 - not
all groups are equal
ANOVA is an omnibus test
you can use use 1 test to compare all the data
assumptions for one-way between-subjects ANOVA
1. each sample is an independent random sample
2. the dependent variable is normally distributed in each
population that is sampled
3. the variance of values of the dependent variable is equal in the
populations being sampled
when/why do you use a repeated measures (within-subjects) ANOVA
when you are comparing the same subjects over a number of levels of
the IV - the groups are dependent, they are the same people tested
multiple times
you would use it if you want to find if there is a difference in
some variable over a period of time or at multiple levels of the variable
power in within-subjects ANOVA vs between-subjects ANOVA
power in a within-subjects (repeated measures) ANOVA is more
powerful: within-subjects gets rid of individual differences
this is because the unexplained variability (error) has gone down,
because you are obtaining results from the same person over time or
with different variables
testing on the same person over time removes sampling variability
and variability caused by the IV
assumptions for within-subjects ANOVA
1. cases represent a random sample of the population of interest and
are independent of each other
2. the dependent variable is normally distributed in the population
for each condition of the study
3. homogeneity of variance - the variances of each of the
populations represented are equal
4. sphericity assumption - the variances of differences between all
pairs of scores over all levels of the factor are identical in the population
contrasts in within-subjects ANOVA
deviation - compares groups to the grand mean
simple - compares 1 group (either first or last) to all other groups
difference - is the time level you chose different from all the time
points that came before it
helmert - (opposite of difference) - same as difference procedure
but starting from the other end
post hoc analyses
after the fact"
- after you get a significant result you have to see where the
significance is- compares all groups to all groups (everyone to
everyone)- atheoretical - not based on theory
use post hoc when you are unsure which group is different from which
group because it compares all groups to all groups
planned comparisons
you decide before you start the analysis which comparisons you are
going to run if the ANOVA comes out significant
- are theoretical: based on theory, logic, practicality
factorial design
a study with more than one independent variable (or factor)
advantage of factorial design: the ability to show interaction effects
ex: a 2x3x4(male vs female) x (major) x (year in school)
main effect
the overall effect of a particular IV
- the main effect of one factor involves using the data for all
levels of the other factor
# of main effects = # of variables (IV)
interaction effect
the effect of one IV depends on or changes with the level of the other
how many main effects are possible in a factorial design?
# of overall effects?
main effects = the same amount as the number of independent variables
overall effects = 2n - 1where n = # of IV's (2 way,
3 way, etc.)- this is how many F values you would get if you ran
an ANOVA
4 parts of discussion section
1. review hypotheses and results
2. discuss the findings in the context of the
existing literature, the implications and
applications of results
3. address the strengths and limitations of your study and provide
suggestions for future research
4. closing
reviewing hypothesis/results in discussion section
#NAME?
discussing context in discussion section
link theory and research from literature review to your findings
- do your results support past research- contradict it?-
what do your results mean in relation to past studies?
discussing implications in discussion section
where the results came from, what the results suggest
why are the findings important? - for the theory and the existing literature
"what does this mean/suggest we need to be aware of
discussing applications in discussion section
real world applications, or theoretical applications - a specific use
to which something is put
"what would you do (or do differently) as a result of this knowledge
discussing limitations in discussion section
internal validity - how well the experiment was done:
were there any experimental errors/issues or measurement
problems- alternative explanations, possible confounds
external validity - generalizability and sample
issues- mundane realism/ ecological validity: the extent to
which findings from experiments/research studies can generalize to
real-life settings
if you discuss a limitation you must propose a way to overcome it in
future research!
discussing strengths in discussion section
what makes your study worthwhile, what makes it better/different from
other studies
discussing future research in discussion section
provide future directions:- what other studies would you do if
you were to continue in this area?
extension of study due to current results:- what changes might
you make in your next study, given what you learned from this one?
address the limitations you suggested and propose a way to fix them
in a future study
closing of discussion section
- reemphasize the study's main strengths- the take home
message/ main point(s) learned- importance of this study in
filling a void in the literature- importance or magnitude of the
overall topic- the "big picture": how can this study
impact behavior and human life- relate to example from introduction
order of research paper
title page
abstract (new page)
introduction (new page)
method
results
discussion
references (new page)
tables (new page)
figures (new page)
appendix (new page)
purpose of an abstract
#NAME?
format of abstract
- goes after the title page but before introduction, on its own page
- 1st line not indented
old version: 120 words maxnew version: 200-250 words max
summarizes all 4 parts of the article:- purpose/thesis (maybe
1 sentence on background info)- quick sentence on methodology/
relevant demographics- was hypothesis supported? - keyword:
"as hypothesized"- future research, implications, applications
can use abbreviations, don't usually include #'s
when is power important?
when you get a non-significant result
if you got a significant result power doesn't matter because you had
to have enough power to obtain that significant result
tables and figures in a research paper
- tables/figures come after the references page when submitting a
research paper
- they are ordered by the way they are mentioned in the text
never include any tables or figures that you did not
discuss/reference in the article!
effect size reference values for eta-squared (ANOVA)
.01 - small.06 - medium.14 - large
eta2 vs partial eta2
eta2 = the proportion of variance
accounted for by each of the main effects, interactions and error in
an ANOVA- usually reported for between-subjects designs, which
have independent cells (different people in each cell)
partial eta2 = the ratio of variance
accounted for by an effect, and that effect + its associated error
variance in an ANOVA- usually reported for within-subjects
(repeated measures) designs where the same people appear in more than
one cell
omega squared
a less bias estimator of the effect size
use for a one-way between-subjects ANOVA
fishing for data
finding significant results and reporting them even when they may
have run analyses wrong (multiple t tests without accounting for type
1 error)
- looking for more information from a data set than it actually contains
data mining
examining your data to see if there is anything else there that you
can talk about in your discussion as future research
cohens d
effect size for t tests
= x1 - x2 / spooled
.2 - small.5 - medium.8 - large
confidence interval
a range of values within which we believe the true population mean falls