Distribution Free Vs. Nondistribution Free Methods
-
Upload
nicolalritter -
Category
Documents
-
view
267 -
download
0
description
Transcript of Distribution Free Vs. Nondistribution Free Methods
![Page 1: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/1.jpg)
1
Distribution Free vs. Non-distribution Free Methods in Factor
Analysis
Nicola Ritter, M.Ed.EPSY 643: Multivariate Methods
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License.
![Page 2: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/2.jpg)
2
Top 5 Take Away Points
1. Extracted factors from a covariance matrix are a function of correlations and standard deviations.
2. Different factors may be extracted based on the matrix of associations selected.
3. Correlational statistics represented in matrices address different questions.
4. Factors are sensitive to the information available in a given correlation statistic.
5. Factors are extracted from a matrix of associations.
![Page 3: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/3.jpg)
3
5. Factors are extracted from a matrix of associations.
• Scores on measured variables are used to compute matrices of bivariate associations.
• i.e. Covariance matrix or correlation matrixEven given only a matrix of associations, all steps
in factor analysis can be completed (except for calculating the factor scores).
VAR1 VAR2 VAR3 VAR4 VAR5 VAR6
VAR1 1
VAR2 1
VAR3 1
VAR4 1
VAR5 1
VAR6 1
![Page 4: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/4.jpg)
4
What are the different types of correlation statistics?
![Page 5: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/5.jpg)
5
4. Factors are sensitive to the information available in a given correlation statistic.
continuous rpb r
rank ρ
categorical Ф rpb
nominal ordinal interval
Bivariate Correlation Coefficients
![Page 6: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/6.jpg)
6
Pearson r Correlation Matrix
• Most commonly used in EFA• Default in most statistical packages
![Page 7: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/7.jpg)
7
Pearson’s r vs. Spearman’s rho
Pearson’s r• Variables are intervally
scaled
Spearman’s ρ• Variables are at least
ordinally scaled.
If the data are intervally scaled, either correlation coefficient could be used.
![Page 8: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/8.jpg)
8
Which correlation coefficient do we use?
![Page 9: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/9.jpg)
9
Pearson r AssumptionParticipant X x̄� x Y Ybar y xy
1 3 4.0 -1.0 3 33.0 -30.0 30.0
2 4 4.0 0.0 4 33.0 -29.0 0.0
3 5 4.0 1.0 92 33.0 59.0 59.0
Sum 12.00 99.00 89.0
Mean 4.00 33.00
SD 1.00 51.10
COV 44.50
r 0.87
![Page 10: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/10.jpg)
10
Spearman rho Assumption
Participant X x̄� x Y Ybar y xy
1 1 2.0 -1.0 1 2.0 -1.0 1.0
2 2 2.0 0.0 2 2.0 0.0 0.0
3 3 2.0 1.0 3 2.0 1.0 1.0
Sum 6.00 6.00 2.0
Mean 2.00 2.00
SD 1.00 1.00
COV 2.00
r 1.00
![Page 11: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/11.jpg)
11
3. Correlational statistics represented in matrices address different questions.
Spearman’s rho1. Addresses the question,
“How well do the two variables order the cases in exactly the same (or the opposite) order?” (Thompson, 2004, p. 130)
Pearson r1. Addresses the same
question AND2. Addresses the question,
“To what extent do the two variables have the same shape?” (Thompson, 2006, p. 130)
![Page 12: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/12.jpg)
12
2. Different factors may be extracted based on the matrix of associations selected.
I. Pearson r Correlation Matrix II. Spearman’s rho Correlation MatrixIII. Covariance Matrix
Data from Thompson, 2004, Appendix A, ID 001-007 & PER1-PER6
![Page 13: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/13.jpg)
13
Pearson’s r Matrix
PER1 PER2 PER3 PER4 PER5
PER1 1.000 0.716 0.654 0.418 0.205
PER2 0.716 1.000 0.720 0.345 0.226
PER3 0.654 0.720 1.000 0.764 0.812
PER4 0.418 0.345 0.764 1.000 0.858
PER5 0.205 0.226 0.812 0.858 1.000
![Page 14: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/14.jpg)
14
Syntax with Factor Analysis with Pearson’s r Matrix
FACTOR /VARIABLES PER1 PER2 PER3 PER4 PER5 /MISSING LISTWISE /ANALYSIS PER1 PER2 PER3 PER4 PER5 /PRINT INITIAL EXTRACTION ROTATION /CRITERIA MINEIGEN(1) ITERATE(25) /EXTRACTION pc /CRITERIA ITERATE(25) /ROTATION varimax /METHOD=CORRELATION.
![Page 15: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/15.jpg)
15
Output of Factor Analysis with Pearson’s r Matrix
Notes. Principal components extraction, varimax-rotated factor matrix
Factor 1 Factor 2 h²
0.169 0.902 0.841
0.161 0.920 0.872
0.749 0.627 0.955
0.911 0.242 0.888
0.985 0.064 0.974
![Page 16: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/16.jpg)
16
Spearman’s rho Matrix
PER1 PER2 PER3 PER4 PER5
PER1 1.000 0.687 0.647 0.392 0.210
PER2 0.687 1.000 0.732 0.283 0.216
PER3 0.647 0.732 1.000 0.574 0.761
PER4 0.392 0.283 0.574 1.000 0.781
PER5 0.210 0.216 0.761 0.781 1.000
![Page 17: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/17.jpg)
17
Syntax with Factor Analysis Spearman’s rho Matrix
NONPAR CORR /VARIABLES=PER1 PER2 PER3 PER4 PER5 /PRINT=SPEARMAN /MATRIX=OUT(*) /MISSING=LISTWISE .
RECODE rowtype_ ('RHO'='CORR') .EXECUTE .
FACTOR /MATRIX=IN(cor=*) /ANALYSIS PER1 PER2 PER3 PER4 PER5 /PRINT INITAL EXTRACTION ROTATION /CRITERIA MINEIGEN(1) ITERATE(25) /EXTRACTION pc /CRITERIA ITERATE(25) /ROTATION varimax /METHOD=CORRELATION .
![Page 18: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/18.jpg)
18
Output of Factor Analysis with Spearman’s rho Matrix
Notes. Principal components extraction, varimax-rotated factor matrix
Factor 1 Factor 2 h²
0.882 0.159 0.804
0.924 0.117 0.868
0.697 0.644 0.900
0.199 0.881 0.815
0.107 0.969 0.951
![Page 19: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/19.jpg)
19
2. Different factors may be extracted based on the matrix of associations selected.
Pearson’s r Spearman’s rho
Notes. Principal components extraction, varimax-rotated factor matrix
Table 1 Table 2
Factor 1 Factor 2 h²
0.169 0.902 0.841
0.161 0.920 0.872
0.749 0.627 0.955
0.911 0.242 0.888
0.985 0.064 0.974
Factor 1 Factor 2 h²
0.882 0.159 0.804
0.924 0.117 0.868
0.697 0.644 0.900
0.199 0.881 0.815
0.107 0.969 0.951
![Page 20: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/20.jpg)
20
1.Extracted factors from a covariance matrix are a function of correlations and standard deviations.
• Matrix most commonly used in CFA• Covariance is Pearson r with standard deviations
removed
rXY = COVXY / (SDX * SDY)COVXY = rXY * SDX * SDY
• Jointly influenced by:1. Correlation between the two variables2. Variability of the first variable3. Variability of the second variable
Thompson (2004)
![Page 21: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/21.jpg)
21
Syntax with Factor Analysis Covariance Matrix
FACTOR /VARIABLES PER1 PER2 PER3 PER4 PER5 /MISSING LISTWISE /ANALYSIS PER1 PER2 PER3 PER4 PER5 /PRINT INITIAL EXTRACTION ROTATION /CRITERIA MINEIGEN(1) ITERATE(25) /EXTRACTION pc /CRITERIA ITERATE(25) /ROTATION varimax /METHOD=cov.
![Page 22: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/22.jpg)
22
Covariance Matrix
PER1 PER2 PER3 PER4 PER5
PER1 1.905 1.238 1.024 0.667 0.381
PER2 1.238 1.571 1.024 0.500 0.381
PER3 1.024 1.024 1.286 1.000 1.238
PER4 0.667 0.500 1.000 1.333 1.330
PER5 0.381 0.381 1.238 1.333 1.810
![Page 23: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/23.jpg)
23
Output of Factor Analysis with Covariance Matrix
Factor 1 Factor 2 h²
0.163 0.923 0.878
0.174 0.898 0.836
0.760 0.615 0.955
0.900 0.250 0.873
0.990 0.055 0.982
Notes. Principal components extraction, varimax-rotated factor matrix
![Page 24: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/24.jpg)
24
Matrices Sensitivity to Different Aspects of the Data
FA with Pearson r Matrix
FA with Spearman rho Matrix
FA with Covariance Matrix
Factor 1
Factor 2 h²
Factor 1
Factor 2 h²
Factor 1
Factor 2 h²
0.169 0.902 0.841 0.882 0.159 0.804 0.163 0.923 0.878
0.161 0.920 0.872 0.924 0.117 0.868 0.174 0.898 0.836
0.749 0.627 0.955 0.697 0.644 0.900 0.760 0.615 0.955
0.911 0.242 0.888 0.199 0.881 0.815 0.900 0.250 0.873
0.985 0.064 0.974 0.107 0.969 0.951 0.990 0.055 0.982
![Page 25: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/25.jpg)
25
Top 5 Take Away Points
1. Extracted factors from a covariance matrix are a function of correlations and standard deviations.
2. Different factors may be extracted based on the matrix of associations selected.
3. Correlational statistics represented in matrices address different questions.
4. Factors are sensitive to the information available in a given correlation statistic.
5. Factors are extracted from a matrix of associations.
![Page 26: Distribution Free Vs. Nondistribution Free Methods](https://reader036.fdocuments.us/reader036/viewer/2022081518/54c4581b4a7959eb2b8b46fc/html5/thumbnails/26.jpg)
26
References
Gorsuch, R.L. (1983). Factor analysis (2nd ed.). Hillsdale, NJ: Erlbaum.
Thompson, B. (2004). Exploratory and confirmatory factor analysis: Understanding concepts and applications. Washington, DC: American Psychological Association.
Thompson, B. (2006). Foundations of behavioral statistics: An insight-based approach. New York, NY: Guilford.