Use LEFT and RIGHT arrow keys to navigate between flashcards;
Use UP and DOWN arrow keys to flip the card;
H to show hint;
A reads text to speech;
59 Cards in this Set
- Front
- Back
Method
|
- Participants
- Measures - Intervention (treatment) - Procedures - Analysis |
|
Descriptive Statistics
|
describe populations via samples
central tendency: mean, median variation: range, standard deviation, percentages, frequencies |
|
Correlational Statistics
|
Describe the relationship between two or more variables (not groups of people)
bi-variate, multiple regression, factor analysis |
|
comparative
|
compare two or more groups on one or more variables. (Experiment is one type of comparative study)
find the effect of the independent variable on the dependent variable - t-test, anova, ancova, manova, mancova |
|
univariate statistics
|
there is one dependent variable
|
|
multivariate
|
more than one dependent variable
|
|
Parametric statistics
|
the dependent variable is measured with and interval or ratio scale
-t-tests, pearson correlations) interval: scores are continuous, equal interval between numbers (likert scales) ratio: scale has an absolute zero. |
|
Non-parametric statistics
|
the dependent variable is measured with nominal or ordinal scale
- Chi-Square analysis, Spearman rank-order, correlation, and many others |
|
inferential statistics
|
you intend to make inferences from the sample to the population
|
|
descriptive statistics
|
you describe a group but not a population (no generalization intended
|
|
Statistically Significant
|
Given the differences between these groups, we are confident that there are similar differences in the population.
or given the magnitude of this relationship between these variables we are confident that there is no a similar relationship in the population .05 or .01 t=3.56, p,.05 means the 2 groups are significantly different at the .05 level. t=.56, p,.05 means that the relationship is significant at the .05 level. |
|
Type I Error
|
you reject the null hypothesis when it is true
ex. you find a significant difference but it does not really exist |
|
Type II Error
|
you accept the null hypothesis when it is false
ex. you don't find a significant difference or relationship but it really does exist in the population |
|
statistical power
|
you can increase statistical power when you do things to avoid Type II errors
- increase sample size - raise the level of significance - use a one-tailed test |
|
true dichotomy
|
not a dichotomy that you created
ex. male and female |
|
artificial dichotomy
|
created by the researcher
|
|
median split
|
split the data in half, high and low sections (artificial dichotomy
|
|
widespread artificial dichotomy
|
take extremem groups (usually the bottom quarter and the top quarter) get ride of all the data in the middle, more likely to find an effect
|
|
partial correlation
|
the relationship between two variables after controlling for a third variable
the correlation may go down if the 3rd variable does contribute ex. the relationship between race and crime after controlling for SES |
|
multiple regression
|
the extent to which various independent variables predict or explain a dependent variable.
ex. to what extent to wealth, health and being in love predict your level of happiness: get % of each variable |
|
discriminant analysis
|
like multiple regression except that the dependent variable is a person's group membership (ex. male or female)
the question: which variables best discriminates between the groups. outcome will tell you how much variance is between each group. |
|
canonical correlation
|
like multiple regression but with sets of variables. Which set of predictor variables best predicts (or correlates with) which set of criterion variables
outcome would tell you which 2-3 variables link up with 2-3 outcomes ex. predictor variables: diet, smoking, exercise criterion variables: cancer, heart disease, cirrhosis |
|
path analysis
|
predicts and then tests the directional relationships of variables. many arrows. Hypothesis and variables are connected by arrows
es. self-esteem is related to substance abuse. self-esteem --> more assertive----> less substance abuse arrows with "ns" if you don't think its directly related put + and - to show positive and negative influence, usually have many paths put data in computer, if you do a path analysis you will get correlation numbers for each arrow |
|
factor analysis
|
determine how clusters of variables are related to each other. clusters are labeled as factors
often used to construct tests. subtests are factors (quantitative and verbal subscores of the GRE are factors) All items have something in common. ex. factor 1: altruism or social factor factor 2: masochist based on the outcome, you make subtests |
|
t-test
|
a parametric statistic comparing two groups on ONE dependent variable or measure.
- more likely to be significant if the difference between the means are large (numerator) and the SDs are small. the larger the n, the smaller the difference between the means have to be for the difference to be significant |
|
chi-square
|
a non-parametric statistic comparing two groups on one dependent variable or measure. DV is non-continuous. ex. yes or no
significant differences are determined by computing the difference between the expected and the observed frequencies. Chi-square is more likely to be significant if there is a big difference between what is expected and observed. expected: null hypothesis, no difference expected observed: what you actually find |
|
ANOVA (Analysis of Variance)
|
comparing more than two groups on ONE dependent variable
- f-test or ratio Will be significant if the variance between the groups is greater than the variance within in the groups likely to have a difference if you have a big SD |
|
Levels
|
ex. 5 x 3 x 2 ANOVA means that there are3 independent variables. one has 5 levels, one has 3 levels, and one has 2 levels
|
|
fully crossed
|
data for every combination is available
nested: not fully crossed |
|
main effect:
|
for each independent variable there is a main effect, when there is more than one independent variable, there can be interactions, 2-way or 3-way and beyond.
something about a particular factor that is making a differnce |
|
interaction
|
the IVs are not independent they have a complex effect on the DV
(lines are not parallel) |
|
contrasts
|
go pack to t-test to compare two groups at a time
can be apriori or post-hoc (aka data-snooping) |
|
ANCOVA - Anaylsis of Covariance
|
ANOVA (2 or more groups on one dv) controls for the effects of a covariate or covariates in the study.
covariate: when you think there is another variable effecting the outcome, you want to control for this variable. ANCOVA with pre-test as a covariate, adjust the post-test scores on the basis of the pre-test differences between the groups. |
|
MANOVA
|
ANOVA with more than one dependent variable
|
|
MANCOVA
|
MANOVA with one or more covariates
|
|
Experimental Validity
|
the extent that the results of the study mean what they are supposed to mean, or that extraneous or confounding variables are controlled or minimized.
|
|
major ways to minimize threats to validity
|
- control group
- random assignment - pretest/postest -double-blind designs |
|
internal threat: History
|
passage of time causes in change
|
|
internal threat: Maturation
|
the maturation process causes change
have a control group |
|
internal threat: testing
|
the test causes change
|
|
internal threat: Instrumentation
|
the use of different measures between pretest and posttest causes change
|
|
internal threat: regression to the mean
|
the tendency that people move from extremes and move toward the middle
|
|
internal threat: differential selection
|
participants are assigned to treatment groups in a biased way
use random assignment |
|
internal threat: Experimental mortality (attrition)
|
incentives, pre-test info, then compare drop-outs with the people that stayed
randomly assign to treatment groups, make the treatments equally desirable |
|
internal threat: The John Henry effect
|
participants know they are in the control group and "compete" with those who get the "real thing"
|
|
internal threat: Experimental treatment diffusion
|
participants in the control get some benefits from the treatment
avoid having the two groups contact, interview participants after to see if there has been a difusion effect |
|
External threat: unclear description of the treatment
|
not able to generalize..
|
|
External threat: multiple treatment interference
|
participants receive other treatment outside of the study. assign subjects to only one treatment
|
|
External threat: the Hawthorne effect
|
attention alone causes change
minimize special attention given to subjects |
|
External threat: Novelty and disruption effects
|
something in the environment causes the change
|
|
External threat: Pretest/Posttest sensitization
|
pretest/post test causes change
|
|
n of 1 studies
|
case study, major limitation is generalization
monitor habits to get a baseline - there are reactive effects just from monitoring, aware of counting habits . ABA: baseline, intervention, baseline ABCA: baseline, intervention, different intervention, baseline |
|
Aptitude-by-treatment interactions
|
hypotheses about interactions
aptitude: started in education, different teaching methods worked for different people, there is a greater effect if the treatment is tailored to the individual's need. |
|
Results Section
|
- objectively and non-defensively present the results
- relevant statistics - only present the important results in the text, use a table for the rest - present results for each hypothesis you had - for qualitative, state themes and illustrative quotes |
|
Discussion Section
|
brief summary of the main results but say verbally.
refer again to the research question and hypotheses - discuss results in reference to the literature, refer to it to explain your results, if your results suppor or conflict with the literature - Limitations - Implications for practice - Suggestions for future research - Conclusion - be specific and modest, avoid "more research is needed" |
|
Simple random sampling
|
all members of a target population have an equal chance of being selected
|
|
systematic sampling
|
every "nth" person on the list of all members of the target population
|
|
Stratified sampling
|
proportionally select participants from subgroups
(ex. if the target population is 40% male, then select a sample that is 40% male) |
|
cluster sampling
|
break the target population into units. Randomly choose some units and use all participants in those units (e.g., randomly choose 7 NFL football teams from the 20(?). Then select all of the players from those 7 teams).
|