Correlation, Regression and Index Numbers

Correlation

Correlation is a statistical measure that describes the extent to which two variables change together. In simpler terms, it tells us if there is a relationship between two sets of data and how strong that relationship is. For example, we might want to know if there's a correlation between the amount of time a student studies and their exam scores.

The correlation coefficient, denoted by 'r', is a value that ranges from -1 to +1.

• r = +1: Perfect positive correlation. As one variable increases, the other variable increases proportionally.
• r = -1: Perfect negative correlation. As one variable increases, the other variable decreases proportionally.
• r = 0: No linear correlation. There is no discernible linear relationship between the two variables.
• 0 < r < 1: Positive correlation. As one variable increases, the other tends to increase, but not perfectly. The closer 'r' is to 1, the stronger the positive relationship.
• -1 < r < 0: Negative correlation. As one variable increases, the other tends to decrease, but not perfectly. The closer 'r' is to -1, the stronger the negative relationship.

Types of Correlation

Correlation can be classified based on the direction and degree of the relationship between variables.

1. Positive Correlation: When two variables tend to move in the same direction. If one increases, the other also increases, and vice-versa. Example: Height and weight (generally, taller people weigh more).
2. Negative Correlation: When two variables tend to move in opposite directions. If one increases, the other decreases. Example: Speed of a car and time taken to reach a destination (higher speed means less time).
3. Zero Correlation: When there is no apparent relationship between the variables. Example: The number of books read by a person and their shoe size.

Methods of Studying Correlation

There are several methods to measure and analyze correlation.

1. Scatter Diagram Method: This is a graphical method. Pairs of values of the two variables are plotted on a graph. The pattern of the plotted points reveals the type and degree of correlation. If the points cluster around a line sloping upwards, it's positive correlation. If they cluster around a line sloping downwards, it's negative correlation. If the points are scattered randomly, there is no correlation.
2. Karl Pearson's Coefficient of Correlation (r): This is a widely used mathematical method that provides a precise numerical value for the correlation between two variables. It measures the linear relationship between two variables.
3. Spearman's Rank Correlation Coefficient: This method is used when data is ranked or ordinal. It measures the strength and direction of the association between the rankings of two variables.
4. Method of Concurrent Deviations: This method considers the direction of change in variables.

Karl Pearson's Coefficient of Correlation (r)

This coefficient is calculated using the following formula:

$$ r = \frac{\sum (x_i - \bar{x})(y_i - \bar{y})}{\sqrt{\sum (x_i - \bar{x})^2 \sum (y_i - \bar{y})^2}} $$

Where:

• $x_i$ and $y_i$ are the individual data points for the two variables.
• $\bar{x}$ and $\bar{y}$ are the means of the x and y variables, respectively.
• $\sum$ denotes summation.

An alternative formula for calculation, which is often simpler for computation, is:

$$ r = \frac{n \sum xy - (\sum x)(\sum y)}{\sqrt{[n \sum x^2 - (\sum x)^2][n \sum y^2 - (\sum y)^2]}} $$

Where 'n' is the number of paired observations.

Spearman's Rank Correlation Coefficient (ρ or $r_s$)

This is used when we have ranked data or when the relationship is monotonic (consistently increasing or decreasing, not necessarily linear). The formula is:

$$ \rho = 1 - \frac{6 \sum d_i^2}{n(n^2 - 1)} $$

Where:

• $d_i$ is the difference between the ranks of the paired observations.
• $n$ is the number of paired observations.

If there are tied ranks, a correction factor is needed, but for exam purposes, the basic formula is usually sufficient unless specified.

Regression Analysis

Regression analysis is a statistical technique used to estimate the relationship between a dependent variable and one or more independent variables. Unlike correlation, which only indicates the strength and direction of a relationship, regression aims to model this relationship to predict the value of the dependent variable based on the values of the independent variable(s).

The most common type is linear regression, where we assume a linear relationship between the variables.

Simple Linear Regression

In simple linear regression, we model the relationship between two variables: one dependent variable (Y) and one independent variable (X). The model is represented by a linear equation:

$$ Y = a + bX + \epsilon $$

Where:

• $Y$ is the dependent variable (the variable we want to predict).
• $X$ is the independent variable (the variable used for prediction).
• $a$ is the Y-intercept (the value of Y when X is 0).
• $b$ is the slope of the regression line (the change in Y for a one-unit change in X).
• $\epsilon$ is the error term, representing the part of Y that cannot be explained by X.

The goal of regression analysis is to find the values of 'a' and 'b' that best fit the data. This is typically done using the "least squares method," which minimizes the sum of the squared differences between the observed values of Y and the values predicted by the regression line.

Estimating the Regression Coefficients (a and b)

The formulas for the slope (b) and intercept (a) using the least squares method are derived from minimizing the sum of squared errors.

The formula for the slope 'b' is:

$$ b = \frac{\sum (x_i - \bar{x})(y_i - \bar{y})}{\sum (x_i - \bar{x})^2} $$

This can also be written as:

$$ b = r \frac{\sigma_y}{\sigma_x} $$

Where:

• $r$ is the Pearson correlation coefficient.
• $\sigma_y$ is the standard deviation of the Y variable.
• $\sigma_x$ is the standard deviation of the X variable.

The formula for the intercept 'a' is:

$$ a = \bar{y} - b\bar{x} $$

This ensures that the regression line passes through the point $(\bar{x}, \bar{y})$, the means of the data.

Making Predictions

Once the regression equation ($Y = a + bX$) is established, we can use it to predict the value of Y for a given value of X. For example, if we have a regression equation predicting exam scores (Y) based on study hours (X), we can plug in a specific number of study hours to estimate the likely exam score.

It's important to note that regression predictions are estimates and may not be perfectly accurate, especially if X is far outside the range of the data used to create the model.

Multiple Linear Regression

This is an extension of simple linear regression where the dependent variable (Y) is predicted using two or more independent variables ($X_1, X_2, ..., X_k$). The equation takes the form:

$$ Y = a + b_1X_1 + b_2X_2 + ... + b_kX_k + \epsilon $$

Here, $b_1, b_2, ..., b_k$ are the regression coefficients for each independent variable, indicating the change in Y for a one-unit change in that specific $X_i$, assuming all other $X$ variables are held constant.

Interpretation of Regression Coefficients

The slope coefficient 'b' is crucial for interpretation.

• Positive 'b': Indicates that as X increases, Y tends to increase.
• Negative 'b': Indicates that as X increases, Y tends to decrease.
• Magnitude of 'b': Shows the strength of the relationship. A larger absolute value of 'b' suggests a stronger impact of X on Y.

The intercept 'a' represents the predicted value of Y when all independent variables are zero. However, 'a' is often not meaningful in real-world contexts if X=0 is impossible or outside the relevant range of data.

Assumptions of Linear Regression

For the results of linear regression to be reliable, several assumptions should ideally be met:

1. Linearity: The relationship between the independent and dependent variables is linear.
2. Independence: The errors ($\epsilon$) are independent of each other.
3. Homoscedasticity: The errors have constant variance across all levels of the independent variables.
4. Normality: The errors are normally distributed.

Violations of these assumptions can affect the validity of the statistical inferences made from the model.

Index Numbers

An index number is a statistical measure that shows changes in a variable or a group of related variables over time. It is typically expressed as a percentage relative to a base period or base value. Index numbers are widely used to track economic trends, inflation, consumer prices, stock market performance, and much more.

For example, the Consumer Price Index (CPI) tracks the average change over time in the prices paid by urban consumers for a market basket of consumer goods and services.

Types of Index Numbers

Index numbers can be broadly classified based on what they measure.

1. Price Index Numbers: Measure changes in the price level. Examples include the Wholesale Price Index (WPI) and Consumer Price Index (CPI).
2. Quantity Index Numbers: Measure changes in the physical volume or quantity of goods produced, consumed, or traded. Example: Index of Industrial Production (IIP).
3. Value Index Numbers: Measure changes in the total value (price times quantity) of a group of items.
4. Special Purpose Index Numbers: Designed for specific purposes, like stock market indices (e.g., Dow Jones, S&P 500).

Methods of Constructing Index Numbers

Constructing an index number involves selecting items, choosing a base period, and using a formula to aggregate the changes.

1. Simple Aggregate Method: The prices (or quantities) of all items in the current period are summed up and compared to the sum of prices (or quantities) in the base period.

$$ \text{Simple Aggregate Index} = \frac{\sum P_1}{\sum P_0} \times 100 $$

Where $P_1$ is the price in the current period and $P_0$ is the price in the base period. This method is simple but gives equal weight to all items and is sensitive to the units of measurement.

2. Average of Relatives Method: This involves calculating the price relative (or quantity relative) for each item and then averaging these relatives.

$$ \text{Simple Average of Relatives} = \frac{1}{n} \sum \left(\frac{P_1}{P_0} \times 100\right) $$

This method avoids the issue of units but still treats all items equally.

3. Weighted Aggregate Method: This is the most common and preferred method, as it accounts for the relative importance of different items. Weights (like quantities consumed, production levels, or economic value) are assigned to each item.

Weighted Index Numbers - Key Formulas

Several weighted index number formulas exist, each with its own advantages.

1. Laspeyres' Index: Uses quantities from the base period ($Q_0$) as weights. It measures the change in cost of purchasing the base period's basket of goods in the current period.

$$ L = \frac{\sum P_1 Q_0}{\sum P_0 Q_0} \times 100 $$

This index tends to overstate inflation because it doesn't account for substitution effects (consumers might switch to cheaper alternatives when prices rise).

2. Paasche's Index: Uses quantities from the current period ($Q_1$) as weights. It measures the change in cost of purchasing the current period's basket of goods using current prices, compared to the base period's prices.

$$ P = \frac{\sum P_1 Q_1}{\sum P_0 Q_1} \times 100 $$

This index tends to understate inflation and can be difficult to compute as current quantities are needed for each period.

3. Fisher's Ideal Index: This is the geometric mean of the Laspeyres and Paasche indices. It is considered "ideal" because it satisfies the time reversal test (the index from period A to B multiplied by the index from B to A equals 1) and the factor reversal test.

$$ F = \sqrt{L \times P} = \sqrt{\frac{\sum P_1 Q_0}{\sum P_0 Q_0} \times \frac{\sum P_1 Q_1}{\sum P_0 Q_1}} \times 100 $$

4. Dorbish and Bowley's Index: This is the arithmetic mean of the Laspeyres and Paasche indices.

$$ DB = \frac{L + P}{2} $$

5. Weighted Average of Relatives: This method assigns weights (w) to the price relatives ($P_1/P_0$).

$$ \text{Weighted Average of Relatives} = \frac{\sum w \left(\frac{P_1}{P_0}\right)}{\sum w} \times 100 $$

If the weights are the base period quantities ($Q_0$), this becomes the Laspeyres index.

Shifting the Base

Sometimes, it is necessary to change the base period of an index number series. This can be done using a simple formula:

$$ \text{New Index} = \frac{\text{Old Index}}{\text{Old Index for New Base Period}} \times 100 $$

For example, if an index has a base year of 2010 and you want to shift the base to 2015, you would divide all index values by the index value for 2015 (which would have been 100 if 2015 was the original base).

Splicing of Index Numbers

Splicing is used when a new index is constructed with a new base period, and you want to continue the old series. It's similar to shifting the base but involves linking two different index series together.

Deflating Index Numbers

When an index number represents value (e.g., sales revenue), it includes the effect of price changes. To find the real change in value (i.e., change in quantity), we "deflate" the value index using a price index.

$$ \text{Real Value Index} = \frac{\text{Value Index}}{\text{Price Index}} \times 100 $$

This process removes the impact of inflation, showing the change in purchasing power or real volume.