###### Covariance of Portfolio Returns Given ...

Covariance between variables can be calculated in two ways. One method is the... **Read More**

A histogram shows the distribution of numerical data in the form of a graph. However, it is very similar to a bar chart, a histogram groups data into intervals. To construct a histogram, you need to establish all the intervals of data, commonly known as bins. The intervals should capture all the data points and also be non-overlapping.

The intervals appear on the horizontal axis, while the **absolute **frequencies appear on the vertical axis. For a histogram with equal intervals in size, a rectangle should be erected over the interval, with its **height** being proportional to the absolute frequency. If intervals are unequal in size, the erected rectangle has an **area **proportional to the absolute frequency of that particular interval. We would have the vertical axis labeled as ‘density’ instead of frequency in such a case. There should be no space between bars to indicate that the intervals are **continuous**.

Consider the previous example of the returns offered by a stock. To bring you up to speed, these were the intervals and the corresponding frequencies:

$$ \begin{array}{c|c} \textbf{Interval} & \textbf{Tally} & \textbf{Frequency} \\ \hline -30\% \leq R_t \leq -20\% & \text{II} & \text{2} \\ -20\% \leq R_t \leq -10\% & \text{I} & \text{1} \\ -10\% \leq R_t \leq 0\% & \text{III} & \text{3} \\ 0\% \leq R_t \leq 10\% & \text{IIIIII} & \text{6} \\ 10\% \leq R_t \leq 20\% & \text{IIIIIII} & \text{7} \\ 20\% \leq R_t \leq 30\% & \text{IIIII} & \text{5} \\ 30\% \leq R_t \leq 40\% & \text{I} & \text{1} \\ \textbf{Total} & \text{} & \textbf{25} \\ \end{array} $$

It is also used to represent the distribution of data graphically. However, it has a major difference when compared to the histogram. Instead of having the class intervals on the horizontal axis clearly showing their upper and lower limits, a frequency polygon uses the **midpoints of the class intervals.**

$$ \text{Midpoint of a class interval} =\text {Lower limit} + \cfrac { (\text{Upper limit} – \text{Lower limit}) }{ 2 } $$

The vertical axis features the absolute frequencies, which are then joined using straight lines and markers.

Going back to the stock return data, we could come up with a frequency polygon.

To come up with the midpoints, we use the formula above. As an example, the midpoint of the interval -30% ≤ R_{t} ≤ -20% is:

$$ \text{Midpoint} = -30 + \cfrac {(-20 – – 30)}{2} = -25 $$

We can calculate the midpoints for the other intervals in a similar manner. The final frequency polygon should look like this:

The frequency polygon is important because it shows the shape of a distribution of data. It can also be very useful when comparing two sets of data side-by-side.

Note: The endpoints touch the X-axis. The vertical scale can also be positioned at the left margin.