Final Flashcards
theory
organization or framework of concepts that permits prediction of data
- organize data, generate prediction, suggest directions for future research
interrater reliability
Is there consistency from rater to rater
- cohen’s kappa
Alternative (parallel form) reliability
Is there consistency between forms of the test?
- correlation coefficient
Inter-item (internal consistency) reliability
Consistency between individual items and the total score?
- cronbach’s alpha
Test-retest reliability
Consistency between scores on tests given in two separate occasions?
- correlation coefficient
Representative reliability
Is the measure consistent across sub populations or groups of people?
- no stat test
Face validity
do the items on the scale appear to measure what you say they do?
panel of experts establish it
Content validity
Is the full content covered?
Criterion validity
How does the measure relate to an already known standard?
- Concurrent: correlate with existing measure of the construct
- Predictive: correlate with other traits that would be associated with what you are trying to measure
Construct validity
How well does the operational definition assess the underlying theoretical construct?
- Discriminative: does the measure differentiate between groups we’d expect to score differently?
- Convergent: do multiple measures of the same construct hang together or operate in consistent ways?
internal validity
can confidently attribute results to the IV (no lurking/confounding variables)
external validity
can confidently generalize results to other people
threats to internal validity - selection bias
other things in common in sample
threats to internal validity - history
Other confounding factors occurring simultaneously
threats to internal validity - maturation
People change over time
threats to internal validity - testing
People get better when tests are repeated
threats to internal validity - instrumentation
Changes to scoring method
threats to internal validity - mortality
attrition, people leaving studies who have similar attributes
threats to internal validity - statistical regression
regression toward the mean, bias only includes the most extreme results (ceiling/floor)
threats to internal validity - diffusion of treatment/contamination
participants in groups communicate
threats to internal and external validity - experimenter expectancy
expectancy affect, experimenter behavior leads subject behavior
threats to external validity - multiple treatment interference
Participants receive additional treatments not part of the study
threats to external validity - Reactive Arrangements (Hawthorne Effect)
Subjects are aware they are being studied
threats to external validity - pretest sensitization
Pretest changes the nature of the treatment, influences performance on posttest
systematic sample
every Kth element, random starting point
stratified sample
population put into subgroups and then randomly selected from each group
cluster sample
researcher obtains a list of clusters not individuals, clusters are randomly sampled
convenience sample
whoever is willing to
purposive sampling
sample is selected from predetermined criteria
snowball sample
ask for referrals
quota sample
try and get a sample that resembles the population (you just go convenience until you meet that criteria)
form of confirmability: triangulation
using multiple methods, sources, and investigators to obtain the same results
3 types of interviews
- informal conversation: discuss a given question, no set topics
- interview guide approach: pre-set topics, no pre-set questions (BEST FOR KIDS)
- standardized open-ended interview: pre-set topics and questions
form of confirmability: persistent observation
observe participants various times
form of confirmability: prolonged engagement
continue to gather data until you’re absolutely sure you know what’s going on
form of confirmability: thick, rich, in-depth description
really tell the story with all the details
form of confirmability: negative case analysis
search for and explain phenomena that do not fit until hypotheses account for all known cases
form of confirmability: audit trail
theoretical memos with detailed descriptions of how you got from data to conclusions
form of confirmability: conceptual saturation
collect data until no new categories are generate
form of confirmability: member checks
present analysis to participants for their confirmation or revision
form of confirmability: peer debriefing
present analyses to other researches (explore inquirer bias)
form of confirmability: explicit documentation
data methods, analysis, and field decisions are all outlined clearly so someone other than you can understand it all
3 aspects of the belmont report
- Respect for persons (autonomy, protection)
- Beneficence (harms minimized, benefits maximized)
- Justice (benefits distributed to participants)
Credibility in qualitative research
Does the data and interpretation actually represent correctly the participants and their reality
Qualitative data analysis: open coding
initial pass through the data, locate initial themes and name them
Qualitative data analysis: axial coding
second and additional passes through the data, focusing on themes more than the data
Qualitative data analysis: selective coding
look selectively for cases that illustrate each theme
sampling in qualitative research
non-probability, produces theory, but usually isn’t generalizable to the population
key informants in qualitative research
someone who has access to the population you want to research
main instrument in qualitative research
YOU
level of significance for p-value
alpha, usually < 0.05
when to use t-test
compare the means of two groups to determine whether the corresponding population means are different
- unpaired: groups are independent
- paired: groups are dependent on or related ot each other in some way
when to use ANOVA (analysis of variance)
compares the average scores of two or more groups
when to use correlation coefficient
assess relationship between 2 variables (no specified IV and DV)
when to use bivariate regression
to measure effect of one control variable
when to use multivariate regression
to measure effect of more than one control variable that also affect each other
type 1 error
false positive, significant stats when in reality the relationship is not significant
type 2 error
false negative, insignificant stats when in reality the relationship is significant
types of qualitative studies
case studies, interviews, ethnography
