Professional Documents
Culture Documents
2010
Poch Bunnak
Understanding Correlations
Perfect P f t Negative correlation Perfect Positive correlation
2010
Poch Bunnak
Pearson Correlation, r
r is a measure of linear association. Assumptions & conditions:
Linear relationship between scale variables (must check this) Normal distribution -1 r +1 Symmetric and Unitless Strongly affected by outliers (Use scatter plot to check this) t-test is used to test H0 that the correlation in the p population p is equal q 0 (See ( the formula on page 272). SPSS reports the p-value. Interpretation: the correlation B/w fathers EDU and sons EDU is found to be r = .65, 65 p < .000. 000 This means that there is a very strong positive relationship between father and son in terms of EDU attainment. The higher the EDU attainment of the father, the higher the EDU attainment of th son! the ! The square of r, R2, (or R2 = SSR/SStotal in ANOVA), is called the coefficient of determination and have PRE interpretation.
Poch Bunnak
4
2010
Other Consideration
Use Scatter plot to check the linearity assumption and the presence of outliers
If these assumptions are violated, use Spearman Rho Practice using SPSS
Poch Bunnak
Formula for r
The coefficient of correlation is calculated as:
r= = ( X X )(Y Y ) (n 1)sx sy
[n(X ) (X ) ][n(Y ) (Y ) ]
2 2 2 2
n(XY) (X )(Y )
t = n2
2010
r 1 r2
6
Poch Bunnak
Example p
RQ: Do men who feel confident in one life domain t d to tend t feel f l confident fid t in i other th domain d i of f life? lif ? Data on self-concepts of men on:
Intimate: Hi score = self-confident in intimate relationship Friend: Hi score = self-confident in relationships among friends Common: Hi score = self-confident in use of common-sense reasoning Academic: Hi score = self-confident in use of academic knowledge in reasoning General: G l Hi score = self-confident lf fid in i the h way Rs R live li their h i lives li
2010
Poch Bunnak
Example, p , cont.
Use Lesson 30 Data file 1 for this exercise Check for normal distribution assumption Check for linear association assumption SPSS: Analyze Correlation Bivariate Select all five variables and move to variables box check on Pearson, , two-tailed, , Flag g significant Options (then child means and S D continue) OK or Paste S.D. Paste.
2010
Poch Bunnak
2010
Poch Bunnak
Result: 7 out of 10 correlations are sifnificant, after adjusting for change of alpha error
2010
Poch Bunnak
10
Poch Bunnak
11
Class Practice
A researcher is interested in relating quality of teaching to quality of research. research His research question is whether professors who are good at teaching are also g good researchers. Use Lesson 30 Exercise File 1 to answer this RQ.
2010
Poch Bunnak
12
Note:
** correlation between all variables . CORRELATIONS /VARIABLES=intimate friend common academic with general /PRINT=TWOTAIL NOSIG /STATISTICS DESCRIPTIVES /MISSING=PAIRWISE. ** correlation between variables from different sets: the first 4 with the last one var. var CORRELATIONS /VARIABLES=intimate friend common g academic with general /PRINT=TWOTAIL NOSIG /STATISTICS DESCRIPTIVES /MISSING PAIRWISE. /MISSING=PAIRWISE.
2010
Poch Bunnak
13
Homework
Problem 1
Ha: If HC is likely to be in a populated area as designated by a market, there would be a positive correlation between the distance to HC and the distance to a market. H0: There is no positive correlation Use CSES village data and run a bivariate correlation between these variables. What do you find? Interpret the finding.
Problem P bl 2
Literature has consistently showed that education is one main factor that lowers the fertility in western society. Use CDHS 2000 data to verify if this was true for Cambodian society or not. Is age at marriage a factor, too? Do the above correlations differ by residential location? Interpret your findings.
2010
Poch Bunnak
14