Brainscape's Knowledge GenomeTM


Top Flashcards (Certified)
All Flashcards

EPPP > Stats > Flashcards

Stats Flashcards

1
Q

Tests of structure

A

Factor analysis, principal components analysis, cluster analysis

What is the underlying structure of a construct?

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

Discriminant analysis

A

Predicts group membership

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

When would you use a factorial/two-way vs. split-plot/mixed ANOVA

A

Factorial: 2 IVs, both independent groups (e.g. TX and gender)

Split-plot: 2 IVs, mixed independent and correlated groups (e.g. TX and time)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

Cross sequential research

A

Combination of cross-sectional and longitudinal, use a cohort but study them over time, but for shorter periods than longitudinal

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What impacts statistical power? (Correctly reject null/find efx)

A

Larger sample size
Stronger intervention
Less error
Parametric test
One-tailed test

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

How to tell if data is independent or correlated?

A

Independent:
Random assignment
Based on pre-existing differences, gender

Correlated:
Measured over time, or repeatedly
Matched subjects
Related subjects

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

Assumptions for parametric tests

A

Interval ratio data
Homoscedasticity same spread between groups
Normally distributed data

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

Assumptions for chi-square test

A

Independent observations, so cannot be measured more than once, or over time

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

Multicolliniarity

A

Multiple regression tests, when predictors are highly correlated with each other

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

Formula for z-scores

A

(Raw score - mean)/SD

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

Canonical R & Canonical Analysis

A

Two x’s and two y’s

Canonical R = correlation
Canonical Analysis = prediction/regression

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

Discriminant function analysis

A

Predict a nominal Y - where people fall into categories- from interval/ratio X, special multiple regression equation. For example, college admissions, or pass fail on an exam

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

Log linear analysis

A

Predicting nominal y with nominal X’s

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Path analysis and structural equation modeling

A

Uses correlational techniques to test out causal models

Path analysis test out researcher identified relationships

Structural equation modeling tests out different paths, uses LISREL

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

Test of difference for ordinal/nominal data, or when I/R data violates assumptions of parametric tests

A

Nominal: Chi-square
- 1 IV
- 2+ IV = multiple sample Chi-square
- Correlated data = McNemar

Ordinal, 1 IV
- 1 group = Kolmigorov (single vodka)
- 2 gps, independent = Kol-Smirnov, Mann-Whitney (double vodka)
- 2 gps, correlated = Wilcoxon (oxen are yoked/correlated)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

Scheffe & Tukey vs. Fischer’s LSD

A

Post-hoc ANOVA tests to identify where sig gp difference coming from

S&T Best protection from type 1 error
Fishers LSD best protection from type 2 error

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

Assumptions of bivariate correlations

A
  1. Homoscedasticity
  2. Unrestricted range on both variables
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

Homoscodasticity

A

Equal variability

across entire scatter plot for correlations

Between groups for tests of difference

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

Which correlation coefficient to use?

A
  • both interval ratio: Pearson r
  • both ordinal: spearman’s Rho, Kendall’s Tau
  • interval ratio and dichotomous point biserial/biserial (point for true dichotomy)
  • true and true dichotomous: phi
  • artificial and artificial dichotomous: tetrachoric
  • curvilinear relationship: eta
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

Latent trait model, or item response Theory

A

Item performance is related to the amount of the respondents latent trait. Latent trait models are used to establish a uniform scale of measurement that can be applied to individuals of varying ability and test content of varying difficulty

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

Classical test Theory

A

Total variability in scores can be explained by combination of test reliability and error variability

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

Cluster analysis

A

Looking for naturally occurring subgroups in data without a priori hypothesis (e.g. profiles of individuals)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
23
Q

Relationship of standard error with standard deviation

A

All of the standard errors have a direct relationship with their related standard deviation

24
Q

Standard error of the mean versus measurement versus estimate

A

All measures of average variability

Standard error of the mean: variability of group mean from population mean. SDpop/sqrt(N)

Standard error of the measurement: variability of individual scores measurement error. Formula includes SDx and reliability coefficient Rxx. Range is 0 - SDx

Standard error of the estimate: variability in prediction error. Formula includes SDy (why why why??) And validity coefficient Rxy. Range is 0 - SDy

25
Criterion deficiency and relevance
Criterion relevance: the extent to which the actual criterion (e.g. whatever measure you're using) truly measures the conceptual criterion Criterion deficiency: the actual criterion is deficient in measuring the conceptual criterion Greater deficiency, less relevance
26
Eigenvalue/ characteristic root
Number that tells you how strong a factor is in factor analysis - how much variance is explained
27
Zero order vs. partial vs. semi-partial correlations
Zero order: most basic, just X and y Partial/first order remove effect of third variable Z from both X and y Semi-partial/ part: remove effect of third variable Z from from only one variable, X or y
28
Multiple R
Multivariate test of: Two or more X's correlated with one y
29
Coefficient of determination
In a correlational relationship, the amount of variability in y explained by X Calculated by squaring the correlation coefficient For multiple R, it's called the coefficient of multiple determination (amount of variability in y explained by multiple X's)
30
Stepwise vs. hierarchical regression
Stepwise regression: computer decides which X variable to enter first based on strength of correlation with y Hierarchical regression: researcher has theory-based order of entering the X variables into the regression
31
Factor loadings
Correlation between a variable (EG. Test item or subtest) and the underlying factor Interpret if greater than or equal to .3
32
Orthogonal vs. Oblique rotation
Methods of doing factor analysis Orthogonal rotation: Factors end up uncorrelated Oblique rotation: factors are correlated (EG WAIS)
33
Communality
Amount of variability explained by factors combined. Calculated by squaring and adding Factor loadings
34
Principle components analysis vs. principle factor analysis
Subtypes types of factor analysis Principle components analysis: no a priori hypothesis, factors are empirically derived Principle factor analysis: theoretically derived factors, already know communalities
35
Three sources of error in test construction
1. Content sampling: by chance you know more or less on a test 2. Time sampling: different scores due to the passage of time 3. Test heterogeneity: items tap multiple domains
36
Four factors affecting reliability
1. Number of items: more items, more reliable 2. Homogeneity of items more homogeneous, more reliable. You want the items to be testing a similar thing 3. Range of scores: greater range, more reliability. So you want a heterogeneous sample 4. Ability to guess: more ability to guess, lower reliability. True, false less reliable than multiple choice
37
Four types of reliability
1. Test retest AKA coefficient of stability 2. Parallel forms AKA alternate forms, AKA equivalent forms 3. Internal consistency reliability - split half - Kuder Richardson or coefficient alpha 4. Interrater reliability
38
Test retest reliability
Aka coefficient of stability Same subjects, same test, different time points Main source of error: time
39
Parallel forms reliability
Aka alternate forms Aka equivalent forms Coefficient of equivalence Same subjects, different tests, different time points Main source of error: time and content sampling
40
Internal consistency reliability
Consistency of scores/items within the test Same subjects, same test, administered once 1. Split half reliability - underestimates reliability because fewer items - to correct for this use. Spearman Brown prophecy formula: how much more reliable is test with X number items? - Bad for speeded tests, good for power tests - main source of error: content sampling 2. Kuder- Richardson and Chronbach's coefficient alpha - analyzes all possible ways of splitting tests in half - main source of error: content sampling and test heterogeneity
41
Interrater reliability
Agreement between scorers on a subjectively score test Improves with group discussion, practice and feedback Measures: % agreement, r, Kappa, Yule's Y
42
What do you need to calculate a confidence band?
1. Raw score 2. Standard error of measurement NOT group mean
43
Three types of validity
1. Content validity: skills and knowledge 2. Criterion-related validity: prediction - concurrent validity: X and Y measured at the same time - predictive validity: X and Y measured with delay 3. Construct validity: traits - convergent - divergent
44
Expectancy tables
Probability that a person 's criterion (y) score will fall in a range given their predictor (x) score
45
Taylor-Russell tables
How much improvement will hiring /selection decisions gain when using a certain test 1. Base rate: rate of successful employees without test. MODERATE (.5) 2. Selection ratio: proportion of available openings to applicants. LOW (.1) 3. Incremental validity: amount of improvement from base rate and success rate when you use the test. Optimized when test has good criterion-related validity
46
Item analysis
When developing a predictor test, how you determine which items in a test to keep Factors: 1. item difficulty: proportion of people who got the item right 2. item discrimination: how well the item discriminates between high and low scorers 3. item validity: correlation between item and whole test score 4. ICC 5. Test revision calculate criterion-related validity coefficient with the revised set of items - Cross validate on a new sample, which always results in shrinkage of the criterion-related validity coefficient
47
Item response Theory (IRT)/ latent trait Theory
To what extent an item correlates with an underlying trait Developed from the item characteristic curve (ICC) Used to develop individually tailored adaptive tests
48
Shrinkage
Relates to cross validation when you are developing a predictor test. When you compare the new test with the original test, the criterion related validity coefficient will always be smaller, because the first sample was used to tailor the items
49
Factors affecting criterion-related validity
1. Range of scores: broad range, heterogeneous subjects 2. Reliability of predictor: reliability puts a ceiling on validity 3. Correction for attenuation: formula that tells you if X and Y were perfectly reliable, how much more valid is the instrument? 4. Criterion contamination: with subjectively scored tests, when the raider knows how person did on the predictor, artificially inflates validity
50
Correction for attenuation
Formula that tells you if criterion (X) and predictor (Y) were perfectly reliable, how much more valid is the instrument?
51
Criterion contamination
With subjectively scored criterion tests, when the raider knows how the person did on the predictor test, artificially inflates the validity of the criterion
52
Multi-trait multi-method matrix
Way to test construct validity, including convergent and divergent validity 1. Convergent validity: how much does scores on new test converge with other measures of the same or similar traits? ---> MONO-trait, HETERO-method, HIGH correlation 2. Divergent validity AKA discriminant validity. How much does the new test differentiate between measures of different traits? ---> HETERO-trait, MONO-method, LOW correlation
53
Time-series design
Establish a longitudinal baseline by taking a bunch of measurements over a period of time, then introduce your experimental manipulation and see if the trend changes Threat of History
54
Central limit theorem
Mean of means = population mean A sample size increases distribution of means becomes more normal
55
Item characteristic curve (ICC)
Part of an item analysis, basis of item response Theory (IRT) Graphs that depict individual test items in terms of the percentage of individuals in different ability groups who answered the item correctly For example, on one item, 80% of people in the highest ability group got it correct, 40% in the middle group, etc

EPPP (12 decks)

Brainscape helps you reach your goals faster, through stronger study habits.
© 2025 Bold Learning Solutions. Terms and Conditions