Professional Documents
Culture Documents
1
ACKNOWLEDGMENT
The success and final outcome of this project required a lot of guidance and
assistance from many people and we are extremely fortunate to have got this all
along the completion of our project work. Whatever we have done is only due to
such guidance and assistance and we would not forget to thank them.
We respect and thank SHIPRA MA’AM, AQM professor for giving us an opportunity
to do the project work in and providing us all support and guidance which made us
complete the project on time. We are extremely grateful to her for providing such a
nice support and guidance even though she had busy schedule.
We are thankful to and fortunate enough to get constant encouragement, support
and guidance from all Teaching staffs which helped us and others in successfully
completing our project work. Also, we would like to extend our sincere regards to all
the non-teaching staff for their timely support.
Thank You.
2
INDEX
S.NO TOPIC PAGE NO.
1 INTRODUCTION 4-5
2 TYPES OF CORRELATION 5-6
3 CORRELATION & CASUAL RELATIONSHIPS 6
4 CASE STUDY 7-10
5 PEARSON’S COEFFICIENT 10-13
6 CONCLUSION 14
7 REFERNCES 15
3
INTRODUCTION
Correlational studies are done to look at the linear relationships between pairs of
variables. There are basically three possible results from a correlation study: a
positive correlation, a negative correlation or no correlation. These are described
below. A scatter diagram is often used to visually see the relationship between the
two variables. A scatter diagram matrix can be used to see the relationship between
multiple pairs of variables. This newsletter will demonstrate how to calculate the
correlation coefficient and determine the probability that it is statistically significant.
4
to use regression or correlation analysis now. Otherwise, complete steps 4
through 7.
Divide points on the graph into four quadrants. If there are X points on the graph,
Count X/2 points from top to bottom and draw a horizontal line.
Count X/2 points from left to right and draw a vertical line.
If number of points is odd, draw the line through the middle point.
Count the points in each quadrant. Do not count points on a line.
Add the diagonally opposite quadrants. Find the smaller sum and the total of
points in all quadrants.
A = points in upper left + points in lower right
B = points in upper right + points in lower left
Q = the smaller of A and B
N=A+B
Look up the limit for N on the trend test table.
If Q is less than the limit, the two variables are related.
If Q is greater than or equal to the limit, the pattern could have occurred from
random chance.
POSITIVE CORRELATION
A positive correlation exists between variable X
and variable Y if an increase in X results in an
increase in Y (and vice-versa).
The more cigarettes you smoke, the greater
the chance of lung cancer.
If you are paid by the hour, the more hours
you work, the more pay you receive.
The more time you spend studying, the better
grades you make in school.
NEGATIVE CORRELATION
A negative correlation exists between variable X
and variable Y if a decrease in X results in an
increase in Y (and vice-versa).
The heavier your car is, the lower your gas
mileage is.
The colder it is outside, the higher your
heating bill.
The more time you spend watching TV, the
lower your grades are in school.
5
NO CORRELATION
In this case, a change in X has no impact on Y (and
vice-versa). There is no relationship between the
two variables. For example, the amount of time I
spend watching TV has no impact on your heating
bill.
6
7
SMOKING AND LUNG CANCER
The example involves a study done in England to look for a correlation between
smoking and lung cancer. Twenty-five occupational groups were included in the
study. The X variable is the number of cigarettes smoked per day relative to the
number smoked by all men of the same age. If X = 100, then the men smoked an
average amount; if above 100, the men smoked more than average and if below
100, they smoked less than average.
The Y variable is the mortality ratio for deaths from lung cancer. Again, if Y = 100,
then the morality ratio for deaths from lung cancer for that group was average;
above 100 it was greater than average and below 100 it was less than average.
The data are shown below. We will use this data to do a correlation analysis of
smoking and mortality.
TABLE 1: OCCUPATIONAL GROUPS
Woodworkers 93 113
8
Makers of other products 112 96
Clerical workers 87 79
Sales workers 91 85
SCATTER DIAGRAM
9
It is always best to begin with a scatter diagram of the data to get a visual picture of
the relationship. For more information on scatter diagrams, please see our February
2005 newsletter. A scatter diagram simply plots the Y variable versus the X variable.
The scatter diagram for the smoking data is shown below.
The scatter diagram does appear to show a positive correlation. The mortality rate
appears to increase as the smoking ratio increases. In addition to this visual picture,
it would be nice to have a numerical measurement of this correlation. This is done by
calculating the correlation coefficient.
where Xi and Yi are the individual paired data points, n is the total number of paired
data points, Xbar is the average X value, Ybar is the average Y value, and sx and sy
are the standard deviations of the X a
nd Y values respectively.
If R is positive, then there is a positive correlation between X and Y. What makes R
positive? Look at the numerator in the R equation:
If an X value is above average and the paired Y value is above average, then the
numerator is positive. If an X value is below average and the paired Y value is below
average, you would be multiplying two negative numbers - which again gives a
positive number. So, in a positive correlation, large value of X tend to generate large
values of Y and small values of X tend to generate small values of Y.
If R is negative, then there is a negative correlation between X and Y. What makes R
negative? Again, look at the numerator in the R equation above. If an X value is
below average and the paired Y value is above average, you will be multiplying a
negative number by a positive number - which generates a negative number. If an X
value is above average and the paired Y value is below average, you will be
multiplying a positive number by a negative number - again generating a negative
number. So, in a negative correlation, large values of X tend to generate small
values of Y and vice-versa.
The calculation of the correlation coefficient is straightforward based on the equation
above. The calculations are shown in the table below.
10
Smoking (X) Mortality (Y) (X-Xbar)*(Y-Ybar)
77 84 647
94 128 -168.72
93 113 -39.52
88 104 74.4
102 88 18.48
91 104 59.4
107 86 -94.76
112 96 -118.56
11
87 79 476.4
91 85 285.12
76 60 1317.12
66 51 2139.04
The R value is positive indicating a positive correlation. This confirms what the
scatter diagram shows.
WHEN IS R SIGNIFICANT?
You have your scatter diagram and the value of R. Is the correlation significant? In
other words, does it mean anything to you in the real world? Some people give
guidelines for interpreting R. For example, one website gives these guidelines:
R Strength
The author does note that others may question his boundaries. They are somewhat
arbitrary. It is much better to determine the "p value" to help you determine if the
correlation is significant. A p value is the probability of getting something more
12
extreme than your result, when there is no correlation between the two variables.
The p value is calculated using the t distribution where:
You can then use the t distribution to find the probability associated this value of t for
23 degrees of freedom. In Excel, use the TDIST function. The value of p is then:
p = 0.000057
This is the probability that you would obtain a result more extreme than the 0.716 if
there was no correlation between the two variables. Since this value is so small, you
assume that there is evidence of a correlation between the two variables.
Many researches assume there is a statistically significant correlation if p <= 0.05. If
the value of p>=0.20, there is no correlation. If the value falls between 0.05 and 0.20,
more data are needed.
IMPACT OF OUTLIERS
Outliers in the data (special causes of variation) can sometimes mask the results.
That is why it is important to construct a scatter diagram to look for outliers. You can
also use a Box-Whisker chart to look for outliers. Suppose you have the data set
below and you want to see if there is c orrelation between X and Y.
X Y
4 12
3 10
9 2
12 20
8 4
3 12
You could calculate the correlation coefficient and the p value. If you do that, you get
the following results:
R = 0.116
p = 0.826
Since p is so large, you conclude
there is no correlation between X and
Y and go happily on your way. The
scatter diagram for the two variables
is shown below. Note the outlier in
the upper right corner.
Since this is so far removed from the
rest of the data, it may be an outlier.
13
So, do you just toss the point out and redo the calculations? No, you should not just
assume that is an outlier. You need to find out if there was a special cause of
variation. If you find the reason for the outlier, you may toss the point out. If not, you
probably need to consider taking more data. If you deleted the point from the
calculations, the results are:
R = -0.962
p = 0.009
So, there is a strong, negative correlation. These data were just to demonstrate the
impact of outliers. You normally would not do a correlation analysis with these few
data points. You should always strive to have as much data as possible to do a
correlation analysis.
R-SQUARED AND R
You can multiply R by itself and get R-Squared. This is the term that is often used in
regression analysis. It gives the % of variation in Y that is explained by the variation
in X. In the lung cancer example, R-squared is 0.513. This means that 51.3% of the
variation in lung cancer mortality is explained by the variation in the smoking ratio.
CONCLUSION
From the diagram and the resultant R, we can conclude that the variables has a
strong positive relation .
That is, the amount of cigarettes smoked increases the chance of deaths from lung
cancer.
14
REFERENCES
https://www.spcforexcel.com/knowledge/root-cause-analysis/correlation-
analysis
http://lib.stat.cmu.edu/DASL/
https://pmstudycircle.com/2014/08/what-is-a-scatter-diagram-correlation-
chart/
15