Titan Insurance Case Study

Group 8 Assignment

Team Members
 Vandana
 Robin
 Bishnu
 Pratik
 Kanupriya

Content
Case Study – Titan Insurance Company 3

Data Analysis 3

Observations 6

Five percent significance test 7

Data Available 7

Hypothesis 7

Calculations 7

Conclusion 8

Reservations about the result 8

Type-I, Type-II Errors & Power of the Test Calculations 9

Type-I Error 9

Type-II Error 9

Power of the Test 10

Appendix A 12

Appendix B 14

Appendix C 15

References 17
Case Study – Titan Insurance Company

The Titan Insurance Company has installed a new incentive payment scheme for its
lift policy sales force. It wants to have an early view of the success or failure of the
new scheme. Indications are that the sales force is selling more policies but sales
always vary in an unpredictable pattern from month to month and it is not clear that
the scheme has made a significant difference.

A questionnaire is submitted to us and we need to formulate statistical findings

based on this questionnaire. The data (Appendix A) under study is for 30
salespersons, selected randomly.

Data Analysis

The following charts show the distribution of data in both the schemes for the given sample
of 30 sales persons.
The table below shows the basic summary and other parameters for the analysis across the
two schemes.

Summary Old Scheme New Scheme

Mean 68033.33 72033.33
Median 67000 74000
Standard Deviation 20455.98 24062.39
Q1 54000 55000
Q3 81500 85750
IQR 27500 30750
Range 82000 90000
 Old and New Scheme Data Anlysis
100000
90000
80000
70000
60000
50000
40000
30000
20000
10000
0
Mean Median Standard Deviation Q1 Q3 IQR Range
Old Scheme New Scheme

Coefficient of correlation (r) = +0.81

Scatter Plot for r=0.81

£140,000
£120,000
£100,000
£80,000
£60,000
£40,000
£20,000
£0
£20,000 £30,000 £40,000 £50,000 £60,000 £70,000 £80,000 £90,000 £100,000 £110,000 £120,000

Test for Normalcy

Tests Old Scheme New Scheme
Mean=median FALSE FALSE
IQR=1.33 SD FALSE FALSE
Range ~ 6 SD FALSE FALSE
 Old Scheme New Scheme
Tests No. of % No. of % Normalcy
Obs share Normalcy Target Obs share Target
Within 1
SD 21 70% 68% 19 63% 68%
Within 2
SD 29 97% 95% 29 97% 95%

Observations
- Given the shape of the distribution histograms, we can assume the data to be
normally distributed even though the data values differ from the theoretical
properties of Normal Distribution.
- The New Scheme histogram is right skewed which means there is a fall in
performance of the salespersons after the implementation of the incentive policy.
Five percent significance test

In this case study, the data under observation is of 30 salespeople taken at two
different periods of time. Also, the population standard deviation is unknown. Hence,
we use Paired t-Test for hypothesis testing.

Data Available

- Significance level (α): 5% (0.05)

- Old scheme sample mean = 68033.33
- New scheme sample mean = 72033.33

Hypothesis
The t-test has two competing hypothesis against which we do the testing, null
hypothesis and alternate hypothesis.

Since, Titan is interested in finding if the new scheme has made any significant
change in the outputs, null hypothesis is set to no difference in the population means
for new and old schemes. The alternate hypothesis is set to new scheme mean being
greater than old one.

µ0= Population mean for the old scheme

µ1= Population mean for new scheme

H0 (null hypothesis): µ0=µ1

H1 (alternate hypothesis): µ1>µ0

Calculations
Using R for the calculations:

Result: -

p-value = 0.06529
 (Image Source: http://www.statskingdom.com/160MeanT2pair.html)

Conclusion

As per the t-test results, p-value is greater than the level of significance (α). Hence, at
95% confidence level, there is not enough evidence that the sales have increased
significantly in the last 4 months under the new scheme (we fail to reject the null
hypothesis). Therefore, we conclude that new scheme has not increased the outputs
significantly.

Reservations about the result

- The box plot for the new scheme shows that the sample’s sales output has
increased after its implementation. However, the t-test results do not provide a
strong evidence that the entire population sales have gone up.

- Also, the data for the new scheme is right skewed. Hence, the sample size of 30 is
not sufficient enough for the entire population.

- The histograms for both the schemes do not follow the proper bell curve.
Type-I, Type-II Errors & Power of the Test
Calculations

Here we have to calculate the Type-I, Type-II errors and the Power of the Test given
that the average output has increased by £5000.

Type-I Error
A Type-I error occurs when we reject the null hypothesis when it is actually true and
should not be rejected. It is also called ‘False Alarm’. The probability of a Type I error
is equivalent to the significance level used for hypothesis testing given that null
hypothesis is true. In this case study, the value of level of significance (α) is 5%. Hence,
the probability of committing Type-I error is 5%.

Type-II Error
A Type-II error occurs when we fail to reject the null hypothesis when it is actually
false and should be rejected. The probability of a Type II error is represented by β.

Data available
Level of significance (α) = 5% (0.05)
No. of samples (n) = 30
Degrees of freedom = 29
Sdsample = 14081.05 (R code: sd(New.Scheme – Old.Scheme)
µ = 5000
X́ = 4000 (Rcode: mean(New.Scheme) - mean(Old.Scheme))

Hypothesis
H0 (null hypothesis): µ <=5000
H1 (alternate hypothesis): µ > 5000 (As mentioned in the problem statement)

Calculations
tcritical = 1.699 (As per the t-table reading)
tstat = -0.3889 (( X́ - µ)/(sd/ √ n ))

Since we need to calculate β (fail reject null hypothesis), the difference in means of
new and old scheme should be zero. Hence, using µ = 0.
 X́ = 4368.176

Therefore, as long as the difference in means is less than £4368.176, the null
hypothesis cannot be rejected. Given that the average output must increase to
£5000, we can calculate the probability of Type II error.

Therefore, the probability of Type-II error is 40.37973% (~ 40%)

Power of the Test

The power of the test, in general terms, is the probability of making the right decision
when null hypothesis is not correct. So, it is 1 - β.

Hence, Power of the Test is 59.62027% (~ 60%)

Using R to verify
 (image source: http://psychstat3.missouristate.edu/Documents/IntroBook3/sbk20.htm)

µ=0 X́ = 4368 µ = 5000

Appendix A

Sample Data

Output (£000)
Salesperso Old New
n Scheme Scheme
1 57 62
2 103 122
3 59 54
4 75 82
5 84 84
6 73 86
7 35 32
8 110 104
9 44 38
10 82 107
11 67 84
12 64 85
13 78 99
14 53 39
15 41 34
16 39 58
17 80 73
18 87 53
19 73 66
20 65 78
21 28 41
22 62 71
23 49 38
24 84 95
25 63 81
26 77 58
27 67 75
28 101 94
29 91 100
30 50 68
 Salesperson Output Comparison
140000

120000

100000

80000

60000

40000

20000

0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30
Old Scheme New Scheme
Appendix B

Assumptions

1. The new incentive payment scheme for its lift policy salesforce on the measures
adopted and the timelines /dates are not in the scope of this report.
2. Number of policies sold are not considered as a driver for each salesperson output
and only the sum assured is the driver for performance indicator.
3. The methodology adopted by Life Insurance Cos form the basis of this study and
hence is used for reference.
4. The payouts to each sales force consists of a low fixed pay and the bonus on their
outputs. Therefore, this study does not cater to the total payouts.
5. The objective is only sales increase for the company which will compensate for its
operational costs. This may mean to break even or be profitable project for the
company.
6. No other costs form part of this project viz. the sales promotion costs for advertising
etc.
7. The time period of evaluation is restricted to 6 months to break even is assumed to
be true in business sense.
Appendix C

Descriptive Statistic formulas used

Description Mathematical Formula R Code

Commonly referred to as the
Mean > mean(<data>)
"average"

Value that lies in the middle of {(n + 1) ÷ 2}th, where “n” is >
Median
the data when arranged in the number of items in the median(<data>)
ascending order set

Standar
d
The standard deviation measures >sd(<data>)
Deviatio
n the spread of the data about the
mean value.
>quantile(<data
Quartile It is the median of the lower half
>). This returns
1 (Q1) of the data set.
all the quartiles
>quantile(<data
Quartile It is the median of the upper half >)
3 (Q3) of the data set. This returns all
the quartiles
Inter
Quartile It is a measure of variability, Q3- Q1 >IQR(<data>)
Range based on dividing a data set into
(IQR) quartiles
>range(<data>)
This returns the
Range The actual spread of data Max value - min value
min & max
values.
T-Test Formulas used

Mathematical
R Code
Description Formula

Dependent t- t.test(<data1>, <data2> ,

This test is used when the
test for paired paired = TRUE, alternative =
samples are dependent;
samples "greater")
that is, when there is only
one sample that has been
tested twice (repeated
measures) or when there
are two samples that have
been matched or "paired"

tcritical value
It is the value that a test
statistic must exceed in
order for the the null
hypothesis to be rejected
Get α = 1- confidnce
level
Calculate degrees of abs(qt(α, DOF))
freedom = n-1
Use t-Table to get the
tcritical value for the
given α & DOF

A test's power is the power.t.test(n, delta , sd,

Power of the
probability of correctly Power = P{Xbar> Xcritical} sig.level,
test
rejecting the null type, alternative)
hypothesis when it is false

References
https://www.pinterest.com/ofncalculators/probability-statistics-
formulas-reference/

https://en.wikipedia.org/wiki/Student's_t-test