Time Series Analysis - A data analysis technique where time is the storyteller. Lecture Note 1.

Time Series Analysis

Lecture note – 1

In this topic, we will learn different types of data analysis techniques. So before we dive into the topic we need to be clear about a few things-

1. What kind of data are we dealing with here?

2. Why do you need different techniques all of a sudden?

3. What is the main objective or goal of this study?

4. How will we proceed towards our goal?

Let’s start with the first question.

Now, what is time-series data?

When we observe the subject repeatedly over some time points then we get time-series data. For example,

(1) the record of daily blood pressure of a particular patient for one week,

(2)quarterly rice production of a certain state,

(3) daily price of a share of a certain company.

Carefully observe one thing that is, blood pressure reading of the patient at any particular day usually not independent of the reading taken on the previous day. If the patient is improving then his blood pressure will tend towards the normal blood pressure level, if the patient’s health condition is not improving then blood pressure reading will tend to get departed from the normal level. So, usually, there will be a pattern in consecutive blood pressure readings, in this case, we cannot assume that the observations are independent.

Suppose, we have the blood pressure reading of that patient for last three days then we can guess the blood pressure reading of the fourth day because we will get a pattern or idea from the past data(whether the patient is improving or not ) and we know the fourth day’s reading will fall in the same pattern. Thus without having any additional information about patients we can make a prediction about his blood pressure on the basis of the data we have in our hands.

Now suppose we are taking the reading of the temperature of the customers arriving in a shopping mall and we have taken the reading of 100 customers. Does the reading of the 101st customer depend on the data of 100 customers? No, each individual has different health conditions, the temperature level of a person can’t depend on any others’ health condition. Thus in this case all the observations are independent and we cannot guess the temperature of 101-th customers without having any knowledge about his health condition.

In time-series data observations are usually not independent. While for the other kind of data we can assume that the observations are independent. We cannot make the assumption of independence for time series data. Here observations share information with the time when it was taken. That’s why we don’t use the usual methods to deal with time-series data.

Question:

Identify time-series data from the following data types:

1. Height of the students of a class

2. Mortality rates of COVID 19 patients of different states in India.

3. Average weekly recovery rates of COVID 19 patients in West Bengal.

4. Annual wheat production of different states in India in 2019.

5. Annual wheat production of India from the year 1990 to 2019.