Guru Gobind Singh Indraprastha University

(GGSIPU)

Submitted in partial fulfilment of requirement of

Masters in Business Administration (MBA)

LAB FILE
OF
DATA PREPARATIONS AND EXPLORATION

Submitted to: Submitted by:

Dr. Shitika DIVYA
00416619824
 DESCRIPTIVE_STATISTICS

EX 1
Q Given the following series of data on Gender and Height for 8 patients, fill in two frequency
tables one for each Variable, according to the model below. Then add a graph. Then create a
contingency table end describe the relation between Gender and Height using appropriate
statistical summaries.
id Height, cm Gender: 1=M, 2=F
1 165 M
2 157 F
3 168 F
4 178 M
5 171 F
6 182 M
7 182 M
8 153 F

modality Absolute freq. Percent freq. Cumulative freq.

SOLUTION 1
Variable Gender:

For Gender we DON'T compute cumulative frequencies as it is a non-ordered qualitative

variable. We will compute the cumulative frequencies for Height, which is quantitative (and thus
necessarily ordered) and continuous.
For Gender, an appropriate graph is with columns (or bars), i.e. with two separate rectangles one
for each modality, M and F, with height proportional to the percentage. It is in general correct to
use a vertical axis going from 0% to 100% in order to avoid a distorted perception of the
importance of the frequencies.

For Height, we consider a division in classes. Let us assume we know the minimum (140 cm)
and the maximum value (200 cm) (notice: a different choice of the extremes as well as of the
number and width of the classes will lead to results slightly different from the following):

The appropriate graphical representation is the histogram, drawn on a cartesian diagram, putting
the classes on the horizontal axis, and drawing for each class a rectangle with height equal to the
frequency density of the class, so that the area of the rectangle equals the frequency of the class.
(Notice the difference with the column chart used for qualitative variables: rectangles here are
contiguous and their area, not their height, is equal to the frequency)
Contingency table:

To describe the relation between Gender and Height, we can compute separately for calcoliamo
M and F the percentages for each age class: these are also called 'row profiles' or conditional
distributions of Height (conditional on Gender, i.e. "restricted to" a specific gender):
This table suggests that M are taller than F. Let us observe also that for Males the Mode is the
class 170 -| 200, while for Females the Mode is 140 - | 160.

EX 2
Q 186 patients were given a therapy for a certain disease; 122 had therapy A, the other 64 had
therapy B.
In group A, there were 37 responders (patients who had benefit from the therapy). In group B,
there were 32 responders. Which was the best treatment? How can we measure the advantage
with this treatment?
Among responders, how many had treatment B?
SOLUTION 2
We fill-in a table with the data given (grey cells) and we complete the missing cells:

(e.g. 85=122-37, 69=37+32. It is a good exercise to repeat what we read in the table, e.g. "there
were 122 patients in group A, 37 of them responded to treatment. In total, 189 patients were
treated, and 69 of them responded")
Now we compute the appropriate percentages:
The best treatment seems to be treatment B. It is intuitive, and it will be seen in the course, that
to compare the percentages in two groups it is useful to compute the ratio. This measure of
comparison of percentages or probabilities is called Risk Ratio (see Probability. The comparison
of numbers via a ratio is illustrated in the Appendix I, among Prerequisites):
RR=50/30.3=1.65
Thus, B has a response percentage superior by 65% with respect to treatment A.
Among responders, the percent of those who got treatment B was:
32/69=46,4%
This percentage is obtained by considering the column profile, or in other terms the distribution
of Treatment conditional on Response equal to Yes.

EX 3
Q Compute the mean and the median class for the following distribution of the number of nurses
in 23 medical institutes.
nurses n
1 – 10 6
11 – 20 13
21 – 40 4
23

SOLUTION 3
The variable Number of Nurses observed in a sample of 23 medical institutes (statistical units) is
a quantitative discrete variable, which we can treat as if it was continuous since it has many
modalities (the numbers from 1 to 40); in fact, the distribution is described by frequencies in
classes. (Let us remark that usually the classes should be contiguous while here there are gaps in
between: this is due to the discrete nature of the variable. This fact has no consequences in our
exercise but it could be annoying when making a graphical representation. For example, in order
to make a representation using a histogram, e.g. the class 1-10 should be considered as 1 |- 11; if
we adopt this convention in our exercise, the mid-values are different from the ones used below)
To compute the mean, we chose a representative value for each class: we take the value in the
middle, computed as (lower limit + upper limit)/2. The total amount of nurses for each class is
then found as this mid-value times the frequency. The mean is the total amount among the
classes divided by the total sample size (the number of statistical units) which is 23.
To identify which is the median class, or in other terms the class that contains the median, we use
the cumulative frequencies.

Mean=365.5 / 23 = 15.5
Median: modality that occupies the rank 12. Looking at the column of cumulative frequencies, it
belongs to the class 11-20 (in fact, the first class includes only the first 6 units; as another
example, the modality that occupies the 20th rank belongs to the class 21-40)

EX 4
Q We know the values of haemoglobin for 6 patients before and after a course of chemotherapy:
we wish to compute the mean reduction. What is the relation between the latter and the means of
the values "before" and "after"?
before after
13.0 9.4
12.8 11.5
11.0 11.5
13.2 13.1
12.5 10.2
11.9 12.0

SOLUTION 4
The Reduction is the difference between the value Before and the value After; in some case a
reduction is negative, this happens since the variable X (haemoglobin) has actually increased.
We can compute the reduction for each of the six patients (statistical units) and then compute
their average. Another way to compute the mean reduction is by using the property of
LINEARITY: given the variables X and Z, if we apply a linear transform such as
Y=aX+bZ
it is always true that mean(Y)=a·mean(X)+b·mean(Z).
In our exercise, a=1 and b= (-1), and thus the mean of the difference is the difference of the
means: mean (Before-After) = mean (Before)-mean (After). All computations are in the table
below.
Notice: the demonstration of such properties will not be a subject of the course tests, as in
general the theoretical results are not part of the knowledge required to the students. This
exercise is illustrated as a complement to the classes.
Another exercise in this document uses again the property of linearity, in the version:
mean(a+bx) =a+bx. Think of a situation when it could be useful to compute the mean of a+bX
when only a, b and mean(X) are known.

EX 5
Q A certain treatment is used in two different centres, A and B; patients in centre A were 25 and
were on average 54 years old; patients treated in centre B were 62 and had mean age equal to 58
years. What is the overall mean among all patients who got the treatment?
SOLUTION 5
We need to compute a weighted average, i.e. the average of two means (54 and 58) weighted by
the size of the two groups (25 and 62).
overall Mean = (54·25 + 58·62) / (25+62) = 4946 / 87 = 56.85

EX 6
Q Pregnant women (within month 4) who are being followed-up by a nutritionist had weights
(kg) equal to: 64.3; 65.2; 70.0; 54.5; 58.8; 81.5; 61.0; 62.0. What was the mean? and the median?
Do data suggest a strong skewness of the distribution of the Weight?
SOLTUION 6
We can sort the observed values and identify the values that occupies the position (rank) 4 and 5
(since we have 8 statistical units - pregnant women).
A slightly different way of illustrating this same procedure is assigning to each value the
corresponding rank in the following table (in other terms, we avoid to write down the values
sorted, but we need to sort to assign the rank!):

Sum of values = 517.3 Mean = 517.3 / 8 = 64.66

Central values (look at ranks 4 and 5): 62 e 64.3 Median = (62 + 64.3) / 2 = 63.15
Since Mean and Median are not very far from each other, the data don't suggest that the variable
Weight has a strongly skewed distribution.
EX 7
The following data regard 10 male adults; we consider age, value of FEV1 (Forced Espiratory
volume in 1 second) and diastolic pressure. Compute median and standard deviation of the three
variables. Then tell what is the variable with higher variation, using an appropriate measure for
comparisons.
Age FEV1 pressure
25 2.5 85
32 1.8 71
28 1.5 92
21 2.5 80
33 4.5 87
33 2.1 83
34 3.4 70
24 1.2 101
41 2.8 90
26 3.9 83

SOLUTION 7
We have 3 quantitative continuous variables. Mean (arithmetic mean) and standard deviation
provide a synthesis of position and variability. The mean is computed as the sum of all values
divided by 10 (n=10 sample size). Let us apply the "fast" formula for computing the standard
deviation. Computations are reported in the table.
To compare variability of these three variables it is NOT sufficient to look at standard deviations,
that by the way are expressed in different units of measurement and describe variables with
different nature! We must express variability in relative terms with respect to the mean, using the
coefficient of variation. The variable with highest variability is FEV1, 4 times more variable than
Pressure and 2 times more variable than Age (notice that FEV1 had apparently the smallest value
for the standard deviation ...)
EX 8
The Age quartiles in a sample of participants in a clinical trial were respectively 27, 41 e 59.
a) This means that:
o 1 out of 4 was younger than … years o 1
out of 4 was older than … years o 2 out of 4
were between … and … years old o half of
them was more than … years old
b) Additionally, we know that mean and standard deviation were respectively equal to 42 and 12.
Can we say whether the distribution was approximately Normal or not?
c) Which index of position is appropriate to give a synthetic description of the distribution?

SOLUTION 8

Point a):
1 out of 4 was younger than 27 years: this is the definition of first quartile, ¼=25% of observed
values was lower than Q1=27
1 out of 4 was older than 59 years: similarly, this is the definition of third quartile, ¾=75%
of observations were lower than Q3, and the other 25% was larger than Q3=59
2 out of 4 are …. We can claim: "between 0 years and the median 41 years", but also
"between Q1 and Q3" and also "between the median 41 and the maximum age" (but we don0t
know the latter).
Half of them was more than 41 years old: this comes from the definition of the Median.
b) First, we notice that the mean is 42 and it is very close to the median, in fact their
distance (equal to 1) is small (1/12) compared to the standard deviation. Thus, the observed
distribution is rather symmetric. But the Normal is not the only kind of symmetric distribution;
we can go further by looking at the quartiles. In a normal curve the first and third quartile should
be at a specific distance from the mean, given by 0.67 times the standard deviation, thus
0.67·12=8. Thus, if the observed distribution was approximately Normal the first and third
quartiles were expected to be 34 e 50. Our observed quartiles are instead 27 and 59, rather distant
from the ones of a normal distribution with the same mean and standard deviation. Thus, our
distribution is NOT Normal-shaped; it was symmetric but not bell-shaped; it could have been a
distribution with very high tails and few observations in the middle, possibly a distribution with
two modes.
Given what we just remarked, neither the mean nor the median are good indexes to describe the
distribution; if it was bi-modal, we should use the two modes, and if we could recognize the
presence of two subpopulations, we should use the means or the medians of these
subpopulations.

EX 9
Consider the 6 patients with values of haemoglobin before and after chemotherapy of EX 4. We
computed the means: respectively 12.40 and 11.28 - and thus 1.12 for the reduction.
Now compute the standard deviation for the variable Before, After and for the Reduction (Before
- After): does the linearity property hold?
SOLUTION 9
The computation of the standard deviations for Before and After is let for the student; the results
are 0.822912 and 1.313646 respectively. For the Reduction, using the "fast" formula:
The variance is:

and the standard deviation is its squared root: 1.595514

Thus, for the standard deviation linearity does not hold - this is because its calculation requires
computing the power 2 of the values and the squared root, and these operations do not follow
linearity: (a + bx)2 ≠ a2 + bx2
We now give the answer to the question of EX 4: using the linearity for the mean is useful for
example when the values are transformed into another unit of measurements of the kind y = a +
bx.
For example, a mean temperature is expressed in Fahrenheit degrees and we want to express it
in Celsius degrees. This won't be possible for the standard deviation!

EX 10
The weight distribution of a sample of adults with physical inabilities is approximately Normal,
with mean 72 and standard deviation 8. Find an interval of values around the mean such that:
a) includes 95% of the observed values
b) includes almost all observed values (and thus coincides with the range, min-max)
c) includes 50% of the observed values

SOLUTION 10
we use the properties of each Normal distribution.
In an interval given by mean ± 2·st.dev. we have about 95% of the values (more precisely, we
should use 1.96 as a factor instead of 2). This answers to question a). similarly for question b)
we shall compute the interval with radius 3·st.dev. which includes 99.7% of values: a) 72 ± 2·8
= (56,88)
b) 72 ± 3·8 = (48,96)
For the last point, let us remark that an interval around the mean=median which includes 50% of
observations is by definition delimited by the first and third quartile (Q1, Q3), thus we can
compute the limits as:
c) 72 ± 0.67·8 = (66.64,77.36)
 ASSIGNMENT-1 (date 27-8)

1. Generate 100 random nos. between 25 to 100

2. Create a table with the following headers

a. Employee ID
b. Department (H.R, Accounts, Admin)
c. Date of Joining [YYYY-MM-DD]
d. Salary before tax
e. Income tax
i. If salary is less than 2 lakhs, income tax charged should be 5% of salary
ii. If salary is in between 2lakhs to 10 lakhs, income tax charged should be
10% of salary
iii. If salary is more than 10 lakhs, income tax charged should be 30% of
salary

1) Find out the average salaries offered by the company.

2) Also find the Count of employees in each department along with the total income tax
paid by them.

3) It should also highlight the employee ID with maximum earnings/ salary.

 Assignment-2 (date 30-8)

You have decided to use Pivot Table to analyse your data and see how business can be improved.

1. How much in $ did you sell in each city?

2. Per each home appliance item, how many items were sold by your company?
3. Now create a table which shows a breakdown of total revenue from each of the items
sold (in the rows), and the regions each item was sold (in columns).

4. Now, based on your last Pivot Table, apply filter so only Microwave, Oven and
Refrigerators will show in "Item".
Additionally, remove NA from the "Region" field.

5. Now, show the Average Discount % per Sales Person. Apply an external filter that
will show only sales made in Columbia. Sort the results in Ascending Order. Who's
the best sales agent.
6. Create a report detailing the $ Sales by Region, Country and Store, in the following
format:
---------
 Assignment-2 (date 27-9)
SPSS

1.

2.

3.
4.
 EXCEL CASES SPORTS DAY

Q. SPORTSMEN DATA STANDARDIZATION

Q. SPORTSMEN DATA FORMATTING
Q SUMMARIZE THE DATA