Research Design and Statistics Flashcards
Pearson r coefficent is used to
meausure the linear relationship between two continuous variables
Eta coefficient is used to
estimate strength on non-linear relationship between two continuous variables
Spearman rho is used to
measure the relationshipt between two sets of ranked data
Biserial
Used to measure the relationship between one continuous and one artificially-made dichotmous variable
Point Biserial
Used to measure the relationship between one continuous variable and one dichotmous variable
Tetrachoric coefficent
used to measure the relationship between two dichotomous variables
Phi coefficent
used to measure the relationship between two dichotomous variables
Mulitiple predictors and a single criterion =
Multiple regression
Analysis that determines which continous variables discriminant between 2+ naturally occurring groups =
discriminant function analysis
Using multiple predictors to sort individuals into 3+ criteron groups
Multiple discriminant analysis
What factor analysis is determine variables/components that account for total variance in scores
Principle component analysis
Rosenthal Effect or Pygmalion Effect
high expectations lead to increased performance
(e. g., Teachers and gifted students)
* Threat to internal validity*
Demand Effect
Participants guess what answers experimenters want
Threat to External Validity
Hawthorne Effect
Subjects behave different just because they are involved in research
(i.e. lightbulbs experiment)
Threats to External Validity
Threats to Internal Validity
History
Maturation
Selection
Experimenter Bias
History
Threat to internal validity
Any external event that affects scores or status on the dependent variable
(ex. previous Bullying intervention in classroom A)
Maturation
Threat to Internal Validity
Any internal (biological or psychological) change that occurs in subjects while the experiment is in progress and systematically effects DV
(i.e., intellectual development between pre and post)
What are some techniques to control for threats to internal validity?
Random Assignment
Blocking
Matching
ANCOVA
Selection
Threat to Internal Validity
Pre-existing subject factors that account for scores on the DV
(ex. Class A is naturally more intelligent than Class B)
Standard deviation is…
square root of the variance
In an normal distribution, the percent of the population that falls between
-1sd to 1sd
68%
68-95-99 rule
Population that falls between a
1sd–2sd–3sd
on a normal curve
z-score of +3 is equivalent to what percentile rank?
99.9 percentile rank or cutoff of .1%
z-score of +2 is equivalent to what percentile rank?
98 percentile rank or cutoff of 2%
z-score of +1 is equivalent to what percentile rank?
84% PR or cutoff point for 16%
z-score of -1 is equal to what percentile rank/cutoff score?
16% PR or cutoff for the bottom 16%
Z scores are
raw scores stated in standard deviation terms
How do you calculate the standard error of the mean?
SEmean= SD/Square root of N
What is the standard error of the mean?
How far the sample mean can be expected to deviate from the corresponding population mean
Beta means
probability of Type II error
Beta means
probability of Type II error
Power means
rejecting the null when it is indeed false
(avoiding Type II error)
Type II error
retaining a false null
(there was an effect and you missed it)
Type I error
rejecting a true null
(saying there is an effect when there is not)
Assumptions of parametric tests
Normal Distribution
Homogeniety of Variance
Independence of Observations
Which assumption of parametric tests is not robust?
Independence of Observations
Alpha means…
the probabibility of making Type I error
How do you combat threats to internal validity?
Random Assignment to groups
Matching
Blocking
Matching
ensures equivalency in an extraneous variable
Blocking
determines effect of extraneous variably by making it an IV
Examples of non-parametric tests
Chi-squared
Mann-Whitney U
Wilcoxin Matched-Pairs Test
Kruskal Wallis Test
Parametric Tests
t-tests
one way ANOVA
factorial ANOVA
MANOVA
types of T-tests
one-sample t-test
independent samples
correlated samples
One sample t-test
compare a sample mean to a population mean
If you are comparing two means you would use a __________
t-test
You use an Independent samples t-test …
when you want to compare means obtained from two independent samples
You use a correlated t-test…
when you want to compare means of two correlated samples
(pre-post test design)
You use a one-way ANOVA when…
there is one IV with 3+ levels
The F statistic in ANOVA represents
the ratio of between group variance to within group variance
You use a factorial ANOVA when…
you have multiple IV’s and one DV
What a factorial ANOVA yields both significant main and interaction effect what does this mean?
That the main effects do not generalize to all situations (the iV acts differently at different levels of another IV)
When do you use a MANOVA
Analyze data from studies with multiple DV’s (outcomes)
reduces Type I error as you avoid running seperate ANOVA’s for each DV
Chi-square test is used to analyzed what data?
Nominal
A chi-square test compares frequencies of observations under nominal categories to…
frequencies expected under the null hypothesis
Mann-Whitney U is used to compare two IV’s to…
DV measured with rank-order data
Alt. to t-test for independent samples
Wilcoxin Matched Pairs Test is used to….
compare two correlated groups to a DV measure in rank-ordered data
alt. to t-test for correlated samples
When do you use a Kruskal Wallis test
when you want to compare 2+ IV’s on a DV with rank-ordered data
alt. to ANOVA
What three non-parametric tests use rank-ordered DV data
Mann Whitney U (2 independent groups)
Wilcoxin Matched-Pairs (2 corr. groups)
Kruskal Wallis Test (2+ independent groups)
Mann Whitney is like
an independent t-test; only uses 2 independent groups (levels)
Wilcoxin Matched-Pairs test is similar to a…
correlated t-test. Uses two correlated groups
Kruskal Wallis is similar to an…
Anova; uses 2 or more independent groups
What posthoc test protects best against Type I error?
Scheffe’s
What posthoc test should you use for pairwise comparisons?
Tukey
Internal Validity
ability to determine if there is a causal relationship between IV and DV
External Validity
Ability to generalize the results of the study to other people, settings, and conditions
Types of random sampling
Stratified random sampling–take a random sample from each subgroup of the total target population
Cluster Sampling –sample a naturally occuring group of individuals that represent the target population
Ways to increase external validity?
Random selection (random sampling)
Naturalistic Research
Single-Double Blind Research
Counterbalancing
Threats to external validity are…
Selection x Treatment
History x Treatment
Testing x Treatment
Demand Characteristics
Hawthorne
Order effects
T-tests cannot be used if there are more than ____ groups to compare
2
T-test means two
What is an eigenvalue?
how much variability a particular factor is accounting for in the other studied variables
(a factor’s strength or explanatory power)
In a PCA, factors are always_______
uncorrelated or correlated?
uncorrelated/independent
The relationship between predictor variables and criterion in a multiple regression model is known as
multiple R or multiple correlation coefficient
Other Correlational Techniques include
Canonical Correlation
Discriminant Function Analysis
Logistic Regression
Multiple Cutoff
Partial Correlation
Canonical Correlation
multiple criterion and multiple predictor variables
Discriminant Function Analysis
Scores on several variables are combained to predict group membership
(backwards from MR)
Predictor variables in a discriminant function analysis require ________
Differential Validity
–each predictor must have a different correlation with each criterion variable
Logistics Regression
Alternative to discriminant function analysis
- Used when assumptions of homogeniety and normality are not met
- Can use nominal or continous data
- Primarily used in research with dichotomous criterion
Path analysis vs. LISREL: what are the differences?
Types of SEM
one-way causal flow vs. one/two-way
observed variables only vs. latent and observed
When should you use a Contingency coefficient
when both variables are nominal (name)
(ex. categories)
Contigency coefficient vs. phi coefficent
nominal vs. dichotomous variables
Random assignment vs. Random selection
Random assignment is for external validity
Random selection is for internal validity
Trend analysis is used in what type of research design?
Repeated measures
measures the nature of an effect and whether the relationship is linear or non-linear
Trend analysis requires what type of IV and DV variables?
Interval or Ratio (quantitative)
Formula for standard error of the mean
sd/sqrt(N)
A change in the raw score at the middle of a distribution will result in [greater/lesser] change in percentile rank compared to a change in raw score at the end of the distribution
greater
Raising the cutoff score on a predictor will
decrease false positives
decrease true positives
(move vertical line to the right)
Item response theory
is typically applied to develop culture faire tests
Minimum F value in an Anova is
1
