Measures of Dispersion: Variance, Standard Deviation, and Mean Deviation

In statistics, we often deal with datasets—collections of numbers that represent observations or measurements. When we analyze a dataset, we're usually interested in its central tendency, which tells us about the typical or average value. Measures like the mean, median, and mode help us understand this central point. However, these measures alone don't tell the whole story. They don't tell us how spread out or concentrated the data points are around the center. This is where measures of dispersion come in.

Measures of dispersion quantify the variability or spread of data points in a dataset. They tell us how much the individual data values differ from each other and from the central value. A dataset with low dispersion is tightly clustered around the mean, while a dataset with high dispersion has data points spread over a wider range. Understanding dispersion is crucial for making informed decisions and drawing accurate conclusions from data, especially in fields like finance, engineering, and scientific research.

Why are Measures of Dispersion Important?

Consider two students who both scored an average of 70 in a mathematics test. Student A scored 70, 70, 70, 70. Student B scored 50, 60, 80, 90. Both have the same mean, but their performance is vastly different. Student A's scores are consistent, while Student B's scores are highly variable. Measures of dispersion help us distinguish between such scenarios.

Key reasons for using measures of dispersion include:

• Understanding Data Variability: To know how spread out or clustered the data is.
• Comparing Datasets: To compare the consistency or variability of two or more datasets.
• Assessing Reliability: To gauge the reliability of central tendency measures. A mean is more representative of data with low dispersion.
• Risk Assessment: In finance, higher dispersion in stock prices indicates higher risk.

Types of Measures of Dispersion

There are several ways to measure dispersion. The most common ones are:

• Range
• Mean Deviation
• Variance
• Standard Deviation
• Quartile Deviation (not covered in detail here, but another common measure)

We will focus on Mean Deviation, Variance, and Standard Deviation in this section.

Mean Deviation

The Mean Deviation (MD) is the average of the absolute differences between each data point and the mean of the dataset. It measures the average distance of each observation from the center of the distribution. Using absolute values ensures that deviations above and below the mean don't cancel each other out.

Formula for Mean Deviation

For an ungrouped dataset $x_1, x_2, \dots, x_n$, the mean deviation about the mean ($\bar{x}$) is calculated as:

$MD(\bar{x}) = \frac{1}{n} \sum_{i=1}^{n} |x_i - \bar{x}|$

Where:

• $n$ is the number of observations.
• $x_i$ is each individual data point.
• $\bar{x}$ is the mean of the dataset.
• $|x_i - \bar{x}|$ is the absolute difference between the data point and the mean.

The Mean Deviation can also be calculated about the median ($M$). In this case, the formula is:

$MD(M) = \frac{1}{n} \sum_{i=1}^{n} |x_i - M|$

Calculation Steps for Mean Deviation (about the mean)

1. Calculate the mean ($\bar{x}$) of the dataset.
2. Find the absolute difference between each data point ($x_i$) and the mean ($\bar{x}$).
3. Sum up all these absolute differences.
4. Divide the sum by the total number of observations ($n$).

Example: Mean Deviation

Calculate the Mean Deviation for the dataset: 2, 4, 6, 8, 10.

Step 1: Calculate the mean ($\bar{x}$)

$\bar{x} = \frac{2 + 4 + 6 + 8 + 10}{5} = \frac{30}{5} = 6$

Step 2: Find absolute differences from the mean

• $|2 - 6| = |-4| = 4$
• $|4 - 6| = |-2| = 2$
• $|6 - 6| = |0| = 0$
• $|8 - 6| = |2| = 2$
• $|10 - 6| = |4| = 4$

Step 3: Sum the absolute differences

Sum = $4 + 2 + 0 + 2 + 4 = 12$

Step 4: Divide by the number of observations

$MD(\bar{x}) = \frac{12}{5} = 2.4$

The Mean Deviation is 2.4. This means, on average, the data points are 2.4 units away from the mean.

Mean Deviation for Grouped Data

For grouped data, where we have class intervals and frequencies, the formula is:

$MD(\bar{x}) = \frac{\sum_{i=1}^{k} f_i |x_i - \bar{x}|}{\sum_{i=1}^{k} f_i}$

Where:

• $k$ is the number of classes.
• $f_i$ is the frequency of the $i$-th class.
• $x_i$ is the midpoint of the $i$-th class.
• $\bar{x}$ is the mean of the grouped data.

Limitations of Mean Deviation

While Mean Deviation is easy to understand, it's not as widely used as Variance and Standard Deviation because:

• It ignores the sign of the deviations, which can sometimes complicate further statistical analysis.
• It doesn't have as strong a theoretical foundation as variance in inferential statistics.

Variance

Variance is a measure of spread that quantifies how far a set of numbers are spread out from their average value. Unlike Mean Deviation, variance uses the squared differences from the mean. Squaring the differences has two main effects: it makes all deviations positive, and it gives more weight to larger deviations.

Formula for Variance

For an ungrouped dataset $x_1, x_2, \dots, x_n$ with mean $\bar{x}$:

Population Variance ($\sigma^2$): If the data represents the entire population.

$\sigma^2 = \frac{1}{N} \sum_{i=1}^{N} (x_i - \mu)^2$

Sample Variance ($s^2$): If the data is a sample from a larger population. We use $n-1$ in the denominator (Bessel's correction) to provide a less biased estimate of the population variance.

$s^2 = \frac{1}{n-1} \sum_{i=1}^{n} (x_i - \bar{x})^2$

Where:

• $N$ is the population size.
• $n$ is the sample size.
• $\mu$ is the population mean.
• $\bar{x}$ is the sample mean.
• $x_i$ is each individual data point.

Calculation Steps for Variance (Sample Variance)

1. Calculate the mean ($\bar{x}$) of the sample data.
2. For each data point ($x_i$), calculate the difference from the mean ($x_i - \bar{x}$).
3. Square each of these differences: $(x_i - \bar{x})^2$.
4. Sum up all the squared differences: $\sum (x_i - \bar{x})^2$.
5. Divide the sum by ($n-1$), where $n$ is the number of data points in the sample.

Example: Variance

Calculate the sample variance for the dataset: 2, 4, 6, 8, 10.

Step 1: Calculate the mean ($\bar{x}$)

$\bar{x} = \frac{2 + 4 + 6 + 8 + 10}{5} = 6$

Step 2: Calculate differences from the mean

• $2 - 6 = -4$
• $4 - 6 = -2$
• $6 - 6 = 0$
• $8 - 6 = 2$
• $10 - 6 = 4$

Step 3: Square the differences

• $(-4)^2 = 16$
• $(-2)^2 = 4$
• $(0)^2 = 0$
• $(2)^2 = 4$
• $(4)^2 = 16$

Step 4: Sum the squared differences

Sum of squares = $16 + 4 + 0 + 4 + 16 = 40$

Step 5: Divide by ($n-1$)

Here, $n=5$. So, $n-1 = 4$.

$s^2 = \frac{40}{4} = 10$

The sample variance is 10.

Variance for Grouped Data

For grouped data, the formula for sample variance is:

$s^2 = \frac{\sum_{i=1}^{k} f_i (x_i - \bar{x})^2}{\sum_{i=1}^{k} f_i - 1}$

Where:

• $k$ is the number of classes.
• $f_i$ is the frequency of the $i$-th class.
• $x_i$ is the midpoint of the $i$-th class.
• $\bar{x}$ is the mean of the grouped data.
• $\sum f_i$ is the total number of observations ($n$).

A computationally simpler formula for variance (especially for manual calculation) is:

$s^2 = \frac{\sum f_i x_i^2 - \frac{(\sum f_i x_i)^2}{n}}{n-1}$

Or for population variance:

$\sigma^2 = \frac{\sum f_i x_i^2}{N} - \mu^2$

Properties of Variance

• Variance is always non-negative ($s^2 \ge 0$).
• If a constant $c$ is added to every value in a dataset, the variance does not change.
• If every value in a dataset is multiplied by a constant $c$, the variance is multiplied by $c^2$.
• Variance is measured in squared units of the original data (e.g., if data is in meters, variance is in square meters). This can make interpretation difficult.

Standard Deviation

Standard Deviation (SD) is the most widely used measure of dispersion. It is simply the square root of the variance. The primary advantage of standard deviation over variance is that it is expressed in the same units as the original data, making it much easier to interpret.

Formula for Standard Deviation

Population Standard Deviation ($\sigma$):

$\sigma = \sqrt{\sigma^2} = \sqrt{\frac{1}{N} \sum_{i=1}^{N} (x_i - \mu)^2}$

Sample Standard Deviation ($s$):

$s = \sqrt{s^2} = \sqrt{\frac{1}{n-1} \sum_{i=1}^{n} (x_i - \bar{x})^2}$

Example: Standard Deviation

Using the same dataset from the variance example: 2, 4, 6, 8, 10. We found the sample variance ($s^2$) to be 10.

Calculate the Standard Deviation:

$s = \sqrt{s^2} = \sqrt{10} \approx 3.16$

The sample standard deviation is approximately 3.16. This means that, on average, the data points in the sample deviate from the sample mean by about 3.16 units. Since the units are the same as the original data (e.g., points, dollars, kilograms), this value is directly interpretable.

Standard Deviation for Grouped Data

For grouped data, you first calculate the variance using the grouped data formula and then take the square root.

$s = \sqrt{\frac{\sum_{i=1}^{k} f_i (x_i - \bar{x})^2}{\sum_{i=1}^{k} f_i - 1}}$

Properties of Standard Deviation

• Standard deviation is always non-negative ($s \ge 0$).
• If a constant $c$ is added to every value, the standard deviation remains unchanged.
• If every value is multiplied by a constant $c$, the standard deviation is multiplied by $|c|$.
• It is measured in the same units as the original data.
• A standard deviation of 0 indicates that all data points are identical.

Interpretation of Standard Deviation

The standard deviation gives us an idea of the typical spread of data.

• Low SD: Data points are clustered closely around the mean.
• High SD: Data points are spread out over a wider range.

For example, in the stock market, a stock with a high standard deviation of its daily returns is considered more volatile and thus riskier than a stock with a low standard deviation.

Chebyshev's Theorem and the Empirical Rule

These are important concepts related to standard deviation that help in understanding data distribution, especially for bell-shaped (normal) distributions.

1. Chebyshev's Theorem

Chebyshev's Theorem applies to *any* distribution, regardless of its shape. It states that for any dataset, the proportion of data values that lie within $k$ standard deviations of the mean is at least $1 - \frac{1}{k^2}$, where $k > 1$.

For example:

• Within $k=2$ standard deviations: At least $1 - \frac{1}{2^2} = 1 - \frac{1}{4} = 0.75$ or 75% of the data.
• Within $k=3$ standard deviations: At least $1 - \frac{1}{3^2} = 1 - \frac{1}{9} \approx 0.889$ or 88.9% of the data.

This theorem provides a lower bound for the data spread.

2. The Empirical Rule (or 68-95-99.7 Rule)

This rule applies specifically to data that follows a **normal distribution** (bell-shaped curve). It provides approximate percentages of data within certain standard deviations of the mean:

• Approximately **68%** of the data falls within 1 standard deviation of the mean ($\bar{x} \pm 1s$).
• Approximately **95%** of the data falls within 2 standard deviations of the mean ($\bar{x} \pm 2s$).
• Approximately **99.7%** of the data falls within 3 standard deviations of the mean ($\bar{x} \pm 3s$).

Comparison of Measures of Dispersion

Let's summarize the key differences and uses:

Measure	Formula (Sample)	Units	Sensitivity to Outliers	Ease of Interpretation	Mathematical Properties
Mean Deviation	$\frac{1}{n} \sum |x_i - \bar{x}|$	Same as data	Less sensitive than variance/SD	Easy	Less robust mathematically
Variance	$s^2 = \frac{\sum (x_i - \bar{x})^2}{n-1}$	Squared units	Sensitive (due to squaring)	Difficult	Good theoretical properties, but squared units are a drawback
Standard Deviation	$s = \sqrt{s^2}$	Same as data	Sensitive (inherited from variance)	Easy	Most widely used, good balance of interpretability and mathematical utility

Choosing the Right Measure

The choice of measure depends on the context:

• Mean Deviation: Useful for a simple, intuitive understanding of average spread, especially when outliers are not a major concern or when you want a less sensitive measure than SD.
• Variance/Standard Deviation: Preferred in most statistical analyses, especially in inferential statistics, hypothesis testing, and modeling. Standard deviation is generally preferred over variance for reporting due to its interpretability in original units.
• Range: Simplest measure (Max - Min), but highly sensitive to outliers and ignores all other data points. Useful for a quick, rough estimate of spread.

Practical Application: Investment Risk

Imagine two investment funds, Fund A and Fund B, with the following annual returns over 5 years:

• Fund A: 10%, 12%, 11%, 13%, 14%
• Fund B: 5%, 15%, 10%, 20%, 15%

Let's calculate the mean and standard deviation for both.

Fund A:

• Mean = (10+12+11+13+14)/5 = 60/5 = 12%
• (Calculations omitted for brevity, but involve finding deviations, squaring them, summing, and dividing by 4, then taking the square root). Let's assume SD(A) ≈ 1.58%

Fund B:

• Mean = (5+15+10+20+15)/5 = 65/5 = 13%
• (Calculations omitted). Let's assume SD(B) ≈ 4.30%

Although Fund B has a slightly higher average return (13% vs 12%), its standard deviation (4.30%) is significantly higher than Fund A's (1.58%). This indicates that Fund B's returns are much more volatile and unpredictable. An investor seeking stability would prefer Fund A, while an investor willing to take on more risk for potentially higher returns might consider Fund B. Standard deviation here serves as a direct measure of investment risk.