Background
Start with data and then look for theorem.
All Model are wrong, but some are useful
1 / 18 Sudheesh K Kattumannil, ISI, Chennai
Background
Start with data and then look for theorem.
All Model are wrong, but some are useful
1 / 18 Sudheesh K Kattumannil, ISI, Chennai
S TATISTICAL M ODELLING : I NTRODUCTION
Sudheesh K Kattumannil, ISI, Chennai
ISI Chennai, 31 January, 2022
2 / 18 Sudheesh K Kattumannil, ISI, Chennai
Motivation
Why studying this course
Choosing an appropriate statistical tool is very much
important in data analysis
Depend on the nature of the variables one has to chose
the tools
3 / 18 Sudheesh K Kattumannil, ISI, Chennai
Motivation
Why studying this course
Choosing an appropriate statistical tool is very much
important in data analysis
Depend on the nature of the variables one has to chose
the tools
3 / 18 Sudheesh K Kattumannil, ISI, Chennai
Motivation
Why studying this course
Choosing an appropriate statistical tool is very much
important in data analysis
Depend on the nature of the variables one has to chose
the tools
3 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 1
A researcher is interested in how variables, such as GRE
(Graduate Record Exam scores), GPA (grade point
average) and prestige of the undergraduate institution,
effect admission into graduate school. The response
variable, admit/don’t admit, is a binary variable.
The data is available in the link
[Link] (”http : //[Link]/stat/data/binary .csv ”)
What is the response variable?
4 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 1
A researcher is interested in how variables, such as GRE
(Graduate Record Exam scores), GPA (grade point
average) and prestige of the undergraduate institution,
effect admission into graduate school. The response
variable, admit/don’t admit, is a binary variable.
The data is available in the link
[Link] (”http : //[Link]/stat/data/binary .csv ”)
What is the response variable?
4 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 2
Consider an experiment in which disease-vector
mosquitoes are exposed to 9 different levels of insecticide,
50 mosquitoes per dose level, with the results shown in the
following table.
Dosage 1 2 3 4 5 6 7 8 9
Mosquitos 50 50 50 50 50 50 50 50 50
Killed 1 2 4 17 26 39 42 48 50
What will be an ideal dosage one can think about while
making mosquito repellents.
At dose level 6.5 what is the expected number of death?
What is the response variable? What is the explanatory
variable?
5 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 2
Consider an experiment in which disease-vector
mosquitoes are exposed to 9 different levels of insecticide,
50 mosquitoes per dose level, with the results shown in the
following table.
Dosage 1 2 3 4 5 6 7 8 9
Mosquitos 50 50 50 50 50 50 50 50 50
Killed 1 2 4 17 26 39 42 48 50
What will be an ideal dosage one can think about while
making mosquito repellents.
At dose level 6.5 what is the expected number of death?
What is the response variable? What is the explanatory
variable?
5 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 2
Consider an experiment in which disease-vector
mosquitoes are exposed to 9 different levels of insecticide,
50 mosquitoes per dose level, with the results shown in the
following table.
Dosage 1 2 3 4 5 6 7 8 9
Mosquitos 50 50 50 50 50 50 50 50 50
Killed 1 2 4 17 26 39 42 48 50
What will be an ideal dosage one can think about while
making mosquito repellents.
At dose level 6.5 what is the expected number of death?
What is the response variable? What is the explanatory
variable?
5 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 2
Consider an experiment in which disease-vector
mosquitoes are exposed to 9 different levels of insecticide,
50 mosquitoes per dose level, with the results shown in the
following table.
Dosage 1 2 3 4 5 6 7 8 9
Mosquitos 50 50 50 50 50 50 50 50 50
Killed 1 2 4 17 26 39 42 48 50
What will be an ideal dosage one can think about while
making mosquito repellents.
At dose level 6.5 what is the expected number of death?
What is the response variable? What is the explanatory
variable?
5 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 2
Consider an experiment in which disease-vector
mosquitoes are exposed to 9 different levels of insecticide,
50 mosquitoes per dose level, with the results shown in the
following table.
Dosage 1 2 3 4 5 6 7 8 9
Mosquitos 50 50 50 50 50 50 50 50 50
Killed 1 2 4 17 26 39 42 48 50
What will be an ideal dosage one can think about while
making mosquito repellents.
At dose level 6.5 what is the expected number of death?
What is the response variable? What is the explanatory
variable?
5 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 3
The number of awards won by students while studying high
school. Predictors of the number of awards earned include
the type of program in which the student was enrolled
(e.g., vocational, general or academic) and the score on
their final exam in math.
6 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 4
The state wildlife biologists want to model how many fish
are being caught by fishermen at a state park.
Visitors are asked how long they stayed, how many people
were in the group, were there children in the group and
how many fish were caught.
Some visitors do not fish, but there is no data on whether a
person fished or not.
Some visitors who did fish did not catch any fish.
Hence, there are excess zeros in the data because of the
people that did not fish.
Zero inflated model
7 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 4
The state wildlife biologists want to model how many fish
are being caught by fishermen at a state park.
Visitors are asked how long they stayed, how many people
were in the group, were there children in the group and
how many fish were caught.
Some visitors do not fish, but there is no data on whether a
person fished or not.
Some visitors who did fish did not catch any fish.
Hence, there are excess zeros in the data because of the
people that did not fish.
Zero inflated model
7 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 4
The state wildlife biologists want to model how many fish
are being caught by fishermen at a state park.
Visitors are asked how long they stayed, how many people
were in the group, were there children in the group and
how many fish were caught.
Some visitors do not fish, but there is no data on whether a
person fished or not.
Some visitors who did fish did not catch any fish.
Hence, there are excess zeros in the data because of the
people that did not fish.
Zero inflated model
7 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 4
The state wildlife biologists want to model how many fish
are being caught by fishermen at a state park.
Visitors are asked how long they stayed, how many people
were in the group, were there children in the group and
how many fish were caught.
Some visitors do not fish, but there is no data on whether a
person fished or not.
Some visitors who did fish did not catch any fish.
Hence, there are excess zeros in the data because of the
people that did not fish.
Zero inflated model
7 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 4
The state wildlife biologists want to model how many fish
are being caught by fishermen at a state park.
Visitors are asked how long they stayed, how many people
were in the group, were there children in the group and
how many fish were caught.
Some visitors do not fish, but there is no data on whether a
person fished or not.
Some visitors who did fish did not catch any fish.
Hence, there are excess zeros in the data because of the
people that did not fish.
Zero inflated model
7 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 4
The state wildlife biologists want to model how many fish
are being caught by fishermen at a state park.
Visitors are asked how long they stayed, how many people
were in the group, were there children in the group and
how many fish were caught.
Some visitors do not fish, but there is no data on whether a
person fished or not.
Some visitors who did fish did not catch any fish.
Hence, there are excess zeros in the data because of the
people that did not fish.
Zero inflated model
7 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 5
Length of a hospital stay of a patients under gone a
surgery
It depended on several factors like age, gender and income
and occupation
8 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 5
Length of a hospital stay of a patients under gone a
surgery
It depended on several factors like age, gender and income
and occupation
8 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 6
Number of cracks in a die which produce metal
components
The life of the die depends on temperature, pressure and
Humidity
9 / 18 Sudheesh K Kattumannil, ISI, Chennai
Example 6
Number of cracks in a die which produce metal
components
The life of the die depends on temperature, pressure and
Humidity
9 / 18 Sudheesh K Kattumannil, ISI, Chennai
Motivation
As mentioned, In data analysis one has to choose an
appropriate statistical tool
Choosing an appropriate statistical tool depends on
several factors
Level of measurements/Type of data
Distribution of the data
The number of group to be analysed
The dependence of the data
10 / 18 Sudheesh K Kattumannil, ISI, Chennai
Motivation
As mentioned, In data analysis one has to choose an
appropriate statistical tool
Choosing an appropriate statistical tool depends on
several factors
Level of measurements/Type of data
Distribution of the data
The number of group to be analysed
The dependence of the data
10 / 18 Sudheesh K Kattumannil, ISI, Chennai
Motivation
As mentioned, In data analysis one has to choose an
appropriate statistical tool
Choosing an appropriate statistical tool depends on
several factors
Level of measurements/Type of data
Distribution of the data
The number of group to be analysed
The dependence of the data
10 / 18 Sudheesh K Kattumannil, ISI, Chennai
Motivation
As mentioned, In data analysis one has to choose an
appropriate statistical tool
Choosing an appropriate statistical tool depends on
several factors
Level of measurements/Type of data
Distribution of the data
The number of group to be analysed
The dependence of the data
10 / 18 Sudheesh K Kattumannil, ISI, Chennai
Motivation
As mentioned, In data analysis one has to choose an
appropriate statistical tool
Choosing an appropriate statistical tool depends on
several factors
Level of measurements/Type of data
Distribution of the data
The number of group to be analysed
The dependence of the data
10 / 18 Sudheesh K Kattumannil, ISI, Chennai
Motivation
As mentioned, In data analysis one has to choose an
appropriate statistical tool
Choosing an appropriate statistical tool depends on
several factors
Level of measurements/Type of data
Distribution of the data
The number of group to be analysed
The dependence of the data
10 / 18 Sudheesh K Kattumannil, ISI, Chennai
Levels of Measurement
Nominal Measurement: Have no meaningful rank among
values.
Ordinal Measurement: Have imprecise differences
between consecutive values, but have a meaningful order
to those values
Interval Scale : Have meaningful distances between
measurements defined, but have no meaningful zero value
defined
Ratio scale: Have a zero value defined and the distances
between different measurements defined
11 / 18 Sudheesh K Kattumannil, ISI, Chennai
Levels of Measurement
Nominal Measurement: Have no meaningful rank among
values.
Ordinal Measurement: Have imprecise differences
between consecutive values, but have a meaningful order
to those values
Interval Scale : Have meaningful distances between
measurements defined, but have no meaningful zero value
defined
Ratio scale: Have a zero value defined and the distances
between different measurements defined
11 / 18 Sudheesh K Kattumannil, ISI, Chennai
Levels of Measurement
Nominal Measurement: Have no meaningful rank among
values.
Ordinal Measurement: Have imprecise differences
between consecutive values, but have a meaningful order
to those values
Interval Scale : Have meaningful distances between
measurements defined, but have no meaningful zero value
defined
Ratio scale: Have a zero value defined and the distances
between different measurements defined
11 / 18 Sudheesh K Kattumannil, ISI, Chennai
Levels of Measurement
Nominal Measurement: Have no meaningful rank among
values.
Ordinal Measurement: Have imprecise differences
between consecutive values, but have a meaningful order
to those values
Interval Scale : Have meaningful distances between
measurements defined, but have no meaningful zero value
defined
Ratio scale: Have a zero value defined and the distances
between different measurements defined
11 / 18 Sudheesh K Kattumannil, ISI, Chennai
Type of data
Categorical/Qualitative:
Nominal
Ordinal
Numeric / Quantitative:
Discrete data
Continuous data: Interval, Ratio.
12 / 18 Sudheesh K Kattumannil, ISI, Chennai
Type of data
Categorical/Qualitative:
Nominal
Ordinal
Numeric / Quantitative:
Discrete data
Continuous data: Interval, Ratio.
12 / 18 Sudheesh K Kattumannil, ISI, Chennai
Choice of Statistical Method
Type of data
Distribution of data
Number of groups to be analysed
Independence of data
13 / 18 Sudheesh K Kattumannil, ISI, Chennai
Choice of Statistical Method
Type of data
Distribution of data
Number of groups to be analysed
Independence of data
13 / 18 Sudheesh K Kattumannil, ISI, Chennai
Choice of Statistical Method
Type of data
Distribution of data
Number of groups to be analysed
Independence of data
13 / 18 Sudheesh K Kattumannil, ISI, Chennai
Choice of Statistical Method
Type of data
Distribution of data
Number of groups to be analysed
Independence of data
13 / 18 Sudheesh K Kattumannil, ISI, Chennai
Choice of Statistical Method
Number of groups to be analysed:
The test to be used depends on whether the data is in one,
two or more groups
The tests for larger number of groups are often extensions
of tests for smaller number of groups, the assumptions
being same
14 / 18 Sudheesh K Kattumannil, ISI, Chennai
Choice of Statistical Method
Number of groups to be analysed:
The test to be used depends on whether the data is in one,
two or more groups
The tests for larger number of groups are often extensions
of tests for smaller number of groups, the assumptions
being same
14 / 18 Sudheesh K Kattumannil, ISI, Chennai
Choice of Statistical Method
Number of groups to be analysed:
The test to be used depends on whether the data is in one,
two or more groups
The tests for larger number of groups are often extensions
of tests for smaller number of groups, the assumptions
being same
14 / 18 Sudheesh K Kattumannil, ISI, Chennai
Type of data
The statistical method to be adopted further depends on
the level of measurement of the variables that have to be
analysed.
For example, we are interested in testing proportions(
median) when the variables are measured in
nominal(ordinal) scale.
15 / 18 Sudheesh K Kattumannil, ISI, Chennai
Type of data
The statistical method to be adopted further depends on
the level of measurement of the variables that have to be
analysed.
For example, we are interested in testing proportions(
median) when the variables are measured in
nominal(ordinal) scale.
15 / 18 Sudheesh K Kattumannil, ISI, Chennai
Distribution of the data
Parametric tests like t and F test assume normality of data
Non-parametric tests do not assume normality and have
fewer assumptions
16 / 18 Sudheesh K Kattumannil, ISI, Chennai
Distribution of the data
Parametric tests like t and F test assume normality of data
Non-parametric tests do not assume normality and have
fewer assumptions
16 / 18 Sudheesh K Kattumannil, ISI, Chennai
Independence of Data
Different statistical tests are used depending on whether
the data are independent, paired or repeated measures.
We use independent t test and paired t test depended on
the independence of the data
For example ANOVA is an extension of t test, while
repeated measure ANOVA is an extension of paired t-test
Independent test: We test the group difference. That is, we
are interested in developing test for one variable and see
whether it is different?
Dependent test: We are interested in variable difference.
We test the difference in two or more variable in same
subject.
17 / 18 Sudheesh K Kattumannil, ISI, Chennai
Independence of Data
Different statistical tests are used depending on whether
the data are independent, paired or repeated measures.
We use independent t test and paired t test depended on
the independence of the data
For example ANOVA is an extension of t test, while
repeated measure ANOVA is an extension of paired t-test
Independent test: We test the group difference. That is, we
are interested in developing test for one variable and see
whether it is different?
Dependent test: We are interested in variable difference.
We test the difference in two or more variable in same
subject.
17 / 18 Sudheesh K Kattumannil, ISI, Chennai
Independence of Data
Different statistical tests are used depending on whether
the data are independent, paired or repeated measures.
We use independent t test and paired t test depended on
the independence of the data
For example ANOVA is an extension of t test, while
repeated measure ANOVA is an extension of paired t-test
Independent test: We test the group difference. That is, we
are interested in developing test for one variable and see
whether it is different?
Dependent test: We are interested in variable difference.
We test the difference in two or more variable in same
subject.
17 / 18 Sudheesh K Kattumannil, ISI, Chennai
Independence of Data
Different statistical tests are used depending on whether
the data are independent, paired or repeated measures.
We use independent t test and paired t test depended on
the independence of the data
For example ANOVA is an extension of t test, while
repeated measure ANOVA is an extension of paired t-test
Independent test: We test the group difference. That is, we
are interested in developing test for one variable and see
whether it is different?
Dependent test: We are interested in variable difference.
We test the difference in two or more variable in same
subject.
17 / 18 Sudheesh K Kattumannil, ISI, Chennai
Independence of Data
Different statistical tests are used depending on whether
the data are independent, paired or repeated measures.
We use independent t test and paired t test depended on
the independence of the data
For example ANOVA is an extension of t test, while
repeated measure ANOVA is an extension of paired t-test
Independent test: We test the group difference. That is, we
are interested in developing test for one variable and see
whether it is different?
Dependent test: We are interested in variable difference.
We test the difference in two or more variable in same
subject.
17 / 18 Sudheesh K Kattumannil, ISI, Chennai
THANK YOU
18 / 18 Sudheesh K Kattumannil, ISI, Chennai