Study Guide content Flashcards
what does it mean to be conservative for a test?
a p-value of 0.01 to be statically significant is more conservative than 0.05. (how much error they will allow into their study).
Factorial design
more than one independent variable
A quasi-experimental design: non equivalent control group
Basically just an experimental design but the researcher had picks who is in the control and experimental groups.
A quasi-experimental design: reversal designs
ABA method
A quasi-experimental design: interrupted time series
Data is being collected at multiple times before and after manipulation is administered. Basically a “longitudinal study” comparing before and after manipulation.
A quasi-experimental design: control series design
create two groups with people of similar characteristics and then you make them the control vs experimental group. SO you can make sure that whatever manipulation you did for one group wouldn’t have happened just in the “wild”
A quasi-experimental design: single subject designs
comparing before and after intervention on one participant
A quasi-experimental design: multiple baseline designs
Introducing the manipulation at different times.
A quasi-experimental design: across subjects, behaviors
Apply the treatment or intervention to all subjects/behaviors simultaneously. and then you Monitor changes in all subjects/behaviors after the intervention.
Program evaluation concept:
constructive/destructive
Adding things one by one or removing things one by one to see which variable is the issue or solution.
Program evaluation concept:
assessment of interventions
-Clearly outline the objectives and goals of the intervention.
- Choose appropriate measuring methods
-track progress.
-Analysis:
-Feedback and Improvement:
what do you use to find the difference between groups?
ANOVA or T-test
what do you use to find the relationship between two groups?
regression and correlation
degrees of freedom
a way you can lie with statistics
-manipulating degrees of freedom
- misinterpreting the results
y-axis
by editing the layout of the y-axis you can lie with statistics
sample bias
lieing with statistics
problem graphics
lying with statistics
research hypothesis
an idea to be tested
Null hypothesis
assumes that there is no effect or relationship. Researchers are trying to reject the null hypothesis in experiments.
effect size
helps us to understand how much difference or relationship is between two variables
power analysis
statistical measure to determine how many participants you want in a study.
Alpha
significance level
Type 1 error
testing positive when you are negative
Type II error
testing negative when you are really positive
Generalization
how well conclusions can generalize to a population.
Meta-analysis
an analysis of a lot of research papers on the same topic
threats to external validity
effects that inhibit generalization.
- sample size
- sample bias
- measurement effects
- regression to the mean
- history effects
- instrumental degradation
- ect.
what to do with outliers
keep them in if they change the outcome too much. Throw them out if it doesn’t really matter.
Sensitivity
Catching the people WITH the condition
Specificity
not wrongly identifyingg that DONT have the conditionn.
Confidence intervals
estimating the likely hood of the sample data to be true about a population.
rate analysis
incidence and prevalence
Incidence
a component of rate analysis:
The number of new cases of a condition that develop in a
population during a defined time. (longitudinal study)
prevalence
a component of rate analysis:
The total number of people in a population with a
condition at a given point in time.(cross sectional)
odds ratio
odds that an event will occur in one group versus the odds that is will occur in another group.
pathological Science
when scientific claims are made without scientific studies to back them up.
What is Content Validity
Does the test cover all of the material
What is Criterion Validity
How well a test predicts or correlated to an established criterion.
What is Construct validity
is it measuring what it is supposed to measure.
what is concurrent validity
Does the test correlate with other tests we have given in the past.
what is convergent vailidty
same as concurrent ( does the test correlate to other tests we have given in the past)
predictive validity
does the test predict future performance.
Meta-analysis
analysis of lots of research papers in one, to draw conclusions.
4 steps to Meta Analysis
1) get on PubMed and find all the research on your wanted topic.
2) make a common metric for the results (example is effect size)
3)code dimensions of the study
4) draw conclusions on the findings.
Longitudinal designs
designs over a long period of time with lots of analysis at different points
cohort designs
follows a group of people over time.
what do you do with missing data
ignore it is best
Quasi experimental design
(If you design it and it makes a better argument use it) We choose who is in the independent and dependent groups. You need to put in the discussion why you decided to do that comparison, and why you did not do randomization.
Two types of replication
exact(illegal with animals) or conceptual(trying to learn something new with the replication)
Pathological science
result from slopy methods and difficult measurements and media.
(Research conclusions not based on good research).
