EXSS273 Exam 4

what is the goal of experimental reasearch?

to determine the effect of the IV on the DV

to identify cause and effect relationships

R =

random assignment of participants to groups

X =

the experiment treatment (IV)

O =

observation/measurement (DV)

delta =

change from pretest to posttest

what are the three categories of experimental deisgn

pre-experimental, quasi-experimental, true experimental

features of pre-experimental

weakest

no random assignment

often no control group

many threats to validy

cannot establish causality

features of quasi-experimental

intermediate

compare groups, but no random assignment

uses pre-existing or convenience groups

better than pre-experimental, but still threats to validity

prone to selection bias and mortality (drop out)

can suggest, but cannot confirm cause and effect

features of true-experimental

strongest

random assignment to groups and a control group

strong control of validty threats (randomization controls for most confounding variables)

strongest evidence for cause and effect

pre-experimental designs

one-shot case study/one-group posttest design

one-group pretest-postest

static group comparison

quasi-experimental designs

pretest-postest non equivalent control group

interrupted time-series design

true experimental designs

posttest only randomized controll group

pretest-postest randomized control trial

one-shot case study

X O

one-group

posttest only

pre-experimental

no control group present

no pretest present

no random assignment

one-group pretest-posttest

O X O

one-group

pretest and postest

pre-experimental

no control group present

static group comparison

X O (experimental)

O (control)

pre-experimental

two groups

no random assignment

post test only

control group present

pretest-postest nonequivalent control group

O X O (experimental)

O O (control)

quasi-experimental

control group present

pretets present

no random assignment

interrupted time series design

O O O X O O O

quasi-experimental

a single group is measured multiple times before and after the treatment

each subject serves as their own control

posttest only randomized control group

R X O (experimental)

R O (control)

true experimental

random assignment to groups

control group present

no pretest

assumes that random assignment made the groups equal, so any difference in the posttest is due to the treatment OR random error

pretest-postest randomized control trial

R O X O (experimental)

R O O (control)

true-experimental

strongest design

random assignment to groups

both groups have pretest postest

descriptive statistics

describing or classifying data from a sample

inferential statistics

drawing conclusions (inferences ) about a population based on sample evidence

frequency distributions

organizes data by showing how often each value occurs

frequency distributions - location

central tendency (mean, median, mode)

frequency distributions - dispersion

spread (range, standard deviation)

frequency distributions - symmetry

skewness

frequency distributions - departure from normality

kurtosis

normal bell curve features

symmetric: equal number of observations on left and right

mean = median = mode (all three measures of central tendency are equal at the center)

68.3% rule: 68.3% of observations fall within one standard deviation of the eman

predictable possibilities: values further from the mean are less common

types of data and their central tendencies

nominal = mode

ordinal = mode, median

interval & ratio = mode, median, mean

one mode

unimodal

two modes

bimodal

tail to right (peak left)

positively skewed

mean > median

tail to left (peak right)

negatively skewed

median > mean

which measure is preferred when data is skewed?

median

where is the mean in a box and whisker plot?

the line in the box

what is variability?

how much scores differ from each other and the mean

range

highest - lowest

deviance

difference between each score and the mean

sum of deviances = 0

variance

average of squared deviations

standard deviation

square root of variance

why do researchers prefer standard deviation over variance?

SD is in the same units as the original data, making it easier to interpret and communicate

variance is in squared units

types of variability used for types of measurement

nominal = nome

ordinal = range

interval & ratio = range, variance, SD

what does greater variability represent?

worse fit of the mean

scores are spread out

mean is less representative

what does lesser variability mean?

better fit of the mean

scores cluster closely around the mean

low variance, low bias (accuracy)

ideal, consistent and accurate

low variance, high bias

consistent but systematically wrong

high variance, low bias

accurate on average but inconsistent

high variance, high bias

inconsistent and inaccurate

statistically discernable difference

p < 0.05

the difference in means is NOT due to chance from a statistical point of view

clinically/practically meaningful difference

the actual difference between means in meaningful in the real world

context dependent: what counts as “meaningful” depends on the field

does this difference matter in practice?

effect size (cohen’s d)

the magnitude/size of the difference between means

how big is the difference?

allows comparison of effects across different measures and studies

cohen’s d = (group1 mean - group2 mean)/average SD

cohen d values

0.00-0.20 = negligible, very small effect

0.20-0.50 = small to medium difference but not easily observed

0.50+ = large, easily observed, meaningful difference

always look at both p-values and effect sizes!

null hypothesis

there is no relationship or euqal relationship between variables

alternative hypothesis

statement of a relationship between variables

non-directional (two-tailed) alternative hypothesis

statement that the means are not equal

directional alternative hypothesis

statement that one mean is higher than the other

steps of hypothesis testing

state the hypothesis (null and alternative must be state BEFORE the study)

select alpha (criterion for rejecting null, usually p < 0.05, set BEFORE the study)

compute test statistics and p-value

make a decision (reject or fail to reject null)

when to reject null?

p < alpha

when to fail to reject null?

p >= alpha

type 1 error

false positive

rejecting null when null is true

convicting innocent person

null symbol

H₀

type 2 error

false negative

failing to reject null when null is false

letting guilty person free

consequences of errors

type 1 is usually more serious in research

Anthony Porter case: wrongly convicted of double homicide and spent 17 years on death row

statistical power

the probability of correctly rejecting a false null hypothesis (avoiding a type 2 error)

larger sample size relation to statistical power

larger sample = higher power = more likely to find true effect

smaller sample size relation to statistical power

smaller sample = lower power = less likely to find true effect

caution about large sample sizes

very large sample sizes can inflate power so much that a very small, meaningless difference becomes statistically significant

always ask “is this difference meaningful?” not just “is it statistically significant?”

what is the total area represented in a normal bell curve?

1

why do we care about normality?

statistical test requirements

predictability

comparability

central limit theorem