Methods And Stats Flashcards
ABAB
Person is first measured over time during baseline -a
Then measured again after tx - b
Then again after tx is taken away -a
Then again after tx - b
cluster sampling
identifying naturally occurring groups or clusters (schools, counties) and then randomly selecting certain of these clusters. Typically, all subjects within the selected clusters are then sampled. Alternatively, subjects may be randomly selected from the clusters
interval recording
time sampling - equal intervals
useful when behaviors have no clear beginning or end (laughing, talking)
event sampling
event recording - observing and recording a behavior each time it occurs
useful for studying behaviors that occur infrequently, have long duration, or leave a permanent record
quasi experimental research
can’t control the assignment of subjects to a treatment group- have to use pre-existing groups
simple random sampling
every member of population has an equal chance of being included in the sample. selection of one member has no effect on selection of another member
reduces bias
stratified random sampling
when population varies in terms of “strata” or characteristics (gender)– divide by the strata and then randomly sample
cluster sampling
use simple or stratified random sampling to select clusters of individuals and then either include all individuals in selected units or randomly selecting individuals from each unit
random selection v random assignment
selection: way subjects are selected from population
assignment: assigned to different levels of IV
systematic error
error due to extraneous variables
random error
error due to random fluctuations in subjects, conditions, etc.
techniques used to control effects of extraneous variables
random assignment to treatment groups
holding extraneous variable constant
matching subjects on the extraneous variable
building the extraneous variable into the study (blocking) (adding it is as an IV)
statistical control (ANCOVA)
internal v external validity
internal: relationship between IV and DV?
external: generalizable?
variables that can threaten internal validity
- maturation
- historical events
- testing , retesting
- instrumentation - changes in measuring devices
- statistical regression - tendency of extreme scores to regress toward the mean
- selection - if assigning to diff tx groups results in diff bt the groups
- attrition
- interactions with selection (e.g. selection and history)
external validity is always limited by __; however, a high degree of __ does not guarantee external validity
internal validity
external validity is affected by these factors
- interactions bt testing and treatment (pretest sensitization)
- interactions bt selection and treatment (e.g. volunteers tend to be more motivated than non volunteers)
- reactivity - respond simply bc their behavior is being observed
- multiple treatment interference - exposing someone to one level of design affected by exposure to previous level
multiple treatment interference - controlled by using a counterbalanced design - diff subj receive levels of IV in a different order
between group design
administer each level to a different group
factorial design
includes 2 or more independent variables
within subjects (repeated measures design)
all levels of IV are administered sequentially to all subjects
single subject designs
AB, ABA, ABAB and multiple baseline design
AB design
single baseline phase (A) and single treatment phase (B)
reversal designs; benefits
ABA, ABAB (withdrawing treatment during baseline phases)
additional control over threats to internal validity
multiple baseline design
use when reversal design is unethical
applying the treatment to different behaviors of the same subject or to same subject in different settings or on different tasks (e.g., use self instruction on homework in 3 diff settings)
nominal scale
unordered categories (gender)
ordinal
ordered categories (more or less of a characteristic ; likert scale- strongly agree)
interval scale
order and equal intervals between points (scores on IQ test)
ratio scale
order, equal intervals, and absolute zero point (e.g., temperature)
frequency of individuals in each category measured on a __ scale; frequency of of aggressive acts, accidents, or hospitalizations represents a __ scale
nominal
ratio
measures of central tendency in positive skew and negative skew
Mo, Md, M
M, Md, Mo
areas under the normal curve
1 SD away: 68%
2 SD away: 95%
3 SD: 99%
SD equation
square root of variance
central limit theorem
- regardless of shape of distribution of normal scores in population, as sample size increases, sampling distribution of the mean approaches a normal distribution
- mean of the sampling distribution of the mean is equal to the population mean
- SD of the sampling distribution of the mean is equal to the population SD / square root of the sample size
standard error of the mean
standard deviation of a sampling distribution of the mean
estimate of the extent to which the mean of any one sample randomly drawn from a population can be expected to vary from the population mean s the result of sampling error — variability d/t sampling error
population SD / square root of population N
alpha
size of rejection region is defined by alpha or level of significance
Type I error
rejects a true null hypothesis (find an effect when there isn’t one)
probability is equal to alpha
type II error
retains a false null hypothesis (don’t find an effect when there is one)
probability is equal to Beta
- alpha is low
- sample size is small
IV is not administered in sufficient intensity
maximize power
- increase alpha
- increase sample size
- increase effects of IV
- minimize error
- use a one-tailed test when appropriate
- use a parametric test
degrees of freedom for t test and chi square
t: N-1
c: Categories - 1
tests for ordinal data
“nonparametric alteranatives” to parametric tests
t-test for independent samples = Mann Whitney
t-test for correlated samples = Wilcoxon
one-way ANOVA = Kruskal-Wallis
effect size
Cohen’s d (how many SD away are the groups) and eta squared (what percentage of viability is accounted for by IV)
practical or clinical significance of results
item difficulty (p)
total number of examinees passing/ total number of examinees (e.g., .50 = 50% passed the item)
item discrimination (D)
discriminates between examinees who obtain high and low scores on the entire test (D = U - L)
e.g., if all examinees in upper group and none in lower group answered item correctly, D = +1.0
discrimination index of .35 or higher is acceptable
item response theory is linked to __; 3 factors associated with this
item characteristic curve;
- level of difficulty (level at which 50% provided correct response)
- ability to discriminate between high and low achievers ((above and below zero)
- probability of guessing correctly (y intercept)
reliability
“truth” “consistency”
proportion of variability in examinees’ obtained scores that is d/t true differences among examinees on that atrtribute
spearman -brown prophecy formula corrects the_
split=half reliability
____ reliability is most thorough method for estimating reliability
alternate forms
internal consistency reliability is not appropriate for
speed tests
___ is a variation of coefficient alpha
Kuder-Richardson
factors that affect reliability
- test length
- range of scores - unrestricted best
- guessing
standard error of measurement
amount of error that can be expected d/t unreliability of test
0 to SDx (test)
lower SD
higher reliability coefficient = smaller SEM
content validity
obtain info about examinee’s familiarity with a particular content
construct validity
determine extent examinee possesses a particular trait
criterion-related validity
predict performance on external criterion
convergent and discriminant validity are ways to assess
construct validity
monotrait - heteromethod coefficients (same trait - different methods) that are large mean
large = convergent validity
heterotrait-monomethod coefficients (diff traits- same method) are small =
discriminant validity
heterotrait-heteromethod coefficients are small =
discriminant validity
orthogonal rotation
uncorrelated;
test’s communality can be calculated by squaring and adding the test’s factor loadings
oblique rotation
correlated
squared factor loading provides a measure of
shared variability
concurrent and predictive validity are associated with
criterion related validity
Standard Error of Estimate
construct confidence interval for criterion score
SDy (smaller) = smaller SEE validity coefficient (larger)= smaller SEE
sensitivity
percent of people in the validation sample who have the disorder and were accurately identified by the predictor
specificity
percent of people who do NOT have the disorder and were accurately identified as NOT Having the disorder
a test’s reliability always places a ceiling on __
validity
cross-validation; validity coefficient tends to __
very important to cross validate a test on another sample
shrink (smaller the initial validation sample, greater the shrinkage)
norm referenced interpretation
comparing an examinee’s test scores to scores obtained by people in normative groups
percentile ranks
standard scores
criterion referenced scores
interpreting based on prespecified standard
percentage
regression equation
expectancy table