BISERIAL AND
POINT BISERIAL
CORRELATION
Presented by –
SULAKSHANA KHATONIYAR
Roll No. 1011021
Research scholar
Dept. of Psychology, Gauhati University
CONTENTS
*What is correlation?
*Methods of computing coefficient of linear
correlation
*Special methods for computing correlation
*Biserial correlation
*Point biserial correlation
*Which one is better and why?
WHAT IS CORRELATION?
*Correlation is a measure of the degree of relationship between two or
more variables.
*Linear correlation is the simplest kind of correlation.
*When the relationship between two sets of scores or variables can be
represented graphically by a straight line, it is known as linear
correlation.
*It can be positive or negative (-1 to +1)
*Zero value indicates absolutely no relationship between two variables.
Positive correlation
Both variables increase or decrease together
Negative correlation
When one variable increases, the other decreases and
vice versa
Zero correlation
No relationship between the variables
Methods of computing * Ordinal level
coefficient of linear * Symbolically
represented by Greek
correlation: letter ρ (rho)
1. Rank Difference
Method
* Charles Spearman
2. Product Moment Method
* Karl Pearson
* Interval or ratio level (continuous)
* Normal distribution
* Symbolically represented by r
SPECIAL METHODS FOR COMPUTING
CORRELATION
Biserial correlation Point biserial correlation
*One variable is continuous One variable is
continuous
A variable that can take on any
numeric value and can be divided into
smaller increments such as fractional
and decimal values.
Eg. – height, weight, time etc.
SPECIAL METHODS FOR COMPUTING
CORRELATION
Biserial correlation Point biserial correlation
*One variable is continuous One variable is
continuous
*Other variable is dichotomous Other variable is dichotomous
A variable that can divided into two
categories
Eg. Yes/ No, True/ False
SPECIAL METHODS FOR COMPUTING
CORRELATION
Biserial correlation Point biserial correlation
*One variable is continuous One variable is continuous
*Other variable is dichotomous Other variable is dichotomous
*Artificial dichotomy Natural dichotomy
Arbitrarily dichotomized Naturally occurring
(Artificial crucial point for division) (Genuine dichotomy, no artificial
crucial point for division)
Examples of artificial and natural dichotomy
Artificial dichotomy Natural dichotomy
* Successful and unsuccessful Male and female
* Athletic and non-athletic Living and dead
* Moral and immoral Color blind and having normal vision
* Poor and not poor Owning a home and not owning a home
* Radical and conservative Being alcoholic and non-
alcoholic
*COMPUTATION
OF BISERIAL COEFFICIENT OF
CORRELATION
rbis =
p = Proportion of cases in one of the categories (higher group) of
dichotomous variable
q = Proportion of cases in the lower group =
Mp = Mean (M) of the values of higher group
Mq = Mean (M) of the values of the lower group
σt = Standard deviation (SD) of the entire group
y = Height of the ordinate of the normal curve separating the portion p
and q
* COMPUTATION OF POINT BISERIAL
COEFFICIENT OF CORRELATION
r p, bis = Or r p, bis =
p = Proportion of cases in one of the categories (higher group) of dichotomous variable
q = Proportion of cases in the lower group = 1 – p
Mp = Mean of the higher group, the first category of the dichotomous variable
Mq = Mean of the values of lower group
Mt = Mean of the entire group
σt = Standard deviation (SD) of the entire group
WHICH ONE IS BETTER AND WHY?
r p,bis is regarded as a better and a much more
dependable statistic than rbis because:
*The point biserial correlation makes no assumptions regarding
the form of distribution in the dichotomized variable where
biserial correlation makes too many assumptions.
*It may be used in regression equation.
*The point biserial r is a product moment r and can be checked
against r. This is usually not possible with rbis.
*Like Pearson r, the range of rp,bis is equal to ±1, but this is not
true for rbis. Due to its range, rp,bis can be easily compared with
other measures of correlation.
*It is always safe to compute rp,bis when we are not sure whether
the dichotomy is natural or artificial. However, the use of rbis is
always restricted to the artificial dichotomy of the
dichotomized variable.
REFERENCES
* Mangal, S. K. (2002). Statistics in Psychology and Education (2nd
ed.). PHI Learning Private Limited.
THANK YOU