Application of Statistics in Psychology

2.1 CORRELATION: PRODUCT MOMENT, PARTIAL CORRELATION, SPECIAL

CORRELATIONS.
Correlation is an expression of the degree and direction of correspondence between two things.
Correlation helps in prediction.
Coefficient of correlation (correlation coefficient) is a number that provides an index of the strength of the
relationship between two things.
Correlation is symbolized as r
Pearson product moment correlation expresses a
1] linear relationship between (only) two variables (X & Y) which are
2] continuous in nature.
The meaning of r is interpreted by its 1] sign & 2] magnitude.
SIGN-
Plus= positive correlation
Minus= negative correlation
None= no correlation/ zero correlation
MAGNITUDE-
Between +1 and -1
The magnitude of correlation is judged by its absolute value. For eg., .90 vs -.90
Positive correlation= two variables simultaneously increase or simultaneously decrease. (direct correlation)
Negative correlation= one variable increase while the other variable decreases. (inverse correlation)
Zero correlation= absolutely no relationship exists between the two variables.
Perfect correlation = If a correlation coefficient has a value of +1 or -1, then the relation between the two
variables being correlated is perfect.
Correlation does not establish causation.
A correlation coefficient is merely an index of the relationship between two variables, not an index of the
causal relationship between two variables. For e.g., strong positive correlation between ice cream sales and
sunburn
Properties of correlation coefficient
1) Range of r
2) The coefficient of determination (r2)
r2 * 100
percentage of variance in Y that is associated with or is accounted for by variance in X
for e.g., if correlation is .60, r2 = .36 and .36*100 = 36. Therefore 36% of variance in Y is accounted for by
X
Remaining variance in Y is called as coefficient of non-determination (k2=(1-r2)
coefficient of alienation is k. if r shows degree of relationship between X and Y, k shows degree of lack of
relationship between X and Y
3) Effect of origin or unit upon correlation coefficient
correlation coefficient does not change when every score in either or both distributions is increased or
multiplied by a constant
4) Correlation and causation
5) Factors influencing correlation coefficient
Greater the variability in the distribution, higher will be the resultant correlation
Taking scores of only extreme groups than random groups produce false high correlation
Heterogenous groups - groups that differ in their mean values (e.g., 2 grades-abstract reasoning and
academic performance)
Addition of an extreme score
Assumptions underlying Pearson’s r
1. Independence among pairs of scores
2. The population of X and population of Y follow normal distribution and the population pair of scores of X
and Y has a normal bivariate distribution
3. X and Y are continuous and linear variables
Types Of Correlations/ Special Correlations
1. Pearson correlation coefficient
Correlation between two continuous and linear variables
2. Point Biserial correlation
Correlation between one continuous variable and one dichotomous variable (true dichotomy)
3. Biserial correlation
Correlation between one continuous variable and one dichotomous variable (artificial dichotomy)
4. Phi correlation
Correlation between two naturally dichotomous variables
5. Tetrachoric correlation
Correlation between two artificially dichotomous variables
6. Spearman rho
Correlation between variables on ordinal scale
Partial Correlation
Partial correlation (rp) is Pearson correlation between two variables X and Y controlled for third variable Z
(rXYZ). It controls the influence of Z on X and Y
Also expressed as correlation between X and Y “controlled for”, “partialed out for”, “held constant for”
variable Z
For e.g., correlation between exercise and weight loss controlled for metabolism rate
Uses Of Partial Correlation
1. useful when theory makes prediction about the role of third variable in correlation
2. useful to statistically control effects of uncontrolled variables
3. useful tool to control the impact of developmental changes by partialling out for the age of participants
4. useful in controlling impact of heterogenous subgroups in correlational analysis
5. useful to control the tendency of social desirability in self-report measures
Semi partial correlation
correlation between X and Y with effect of Z controlled for Y
For e.g., academic achievement and intelligence with conscientiousness controlled for AA
Multiple correlation
Correlation of one variable with multiple other variables