Calculating Summary Statistics

Okay, you've loaded your data and are ready to start examining it. The first step in understanding any dataset is often calculating basic descriptive statistics. These numbers summarize the data's main characteristics, giving you a quick snapshot of its central value and how spread out the values are. Think of them as the signs for your dataset.

We'll focus on two primary types of summary statistics: measures of central tendency and measures of spread (or variability).

Measures of Central Tendency: Finding the Center

Measures of central tendency aim to describe the "typical" or "central" value in your dataset. What single number best represents the entire group? The three most common measures are the mean, median, and mode.

The Mean (Average)

The mean is probably the most familiar measure. It's simply the sum of all the values divided by the number of values. If you have a dataset with $n$ observations represented as $x_1, x_2, ..., x_n$, the sample mean (often denoted as $\bar{x}$) is calculated as:

$$\bar{x} = \frac{\sum_{i=1}^{n} x_i}{n} = \frac{x_1 + x_2 + ... + x_n}{n}$$

Example: Let's say we have the ages of 5 employees: [25, 30, 28, 45, 28]. The sum is $25 + 30 + 28 + 45 + 28 = 156$. The number of values is $n = 5$. The mean age is $\bar{x} = 156 / 5 = 31.2$ years.

The mean uses every value in the dataset, which is good, but it also makes it sensitive to outliers (extremely high or low values). That 45-year-old pulls the average age up. If that value were 85 instead, the mean would jump significantly, even though most employees are much younger.

The Median ($M_e$): The Middle Ground

The median is the middle value when the data is sorted in ascending order. It splits the dataset exactly in half: 50% of the values are below the median, and 50% are above it.

To find the median:

1. Sort the data from smallest to largest.
2. If there's an odd number of values ($n$ is odd), the median is the middle value.
3. If there's an even number of values ($n$ is even), the median is the average of the two middle values.

Example (Odd n): Using the sorted ages [25, 28, 28, 30, 45]. The number of values is $n = 5$ (odd). The middle value is the $(n+1)/2 = (5+1)/2 = 3$rd value. The median ($M_e$) is 28.

Example (Even n): Let's add another age, 22: [22, 25, 28, 28, 30, 45]. The number of values is $n = 6$ (even). The middle two values are the $n/2 = 6/2 = 3$rd and the $(n/2)+1 = 4$th values. These are 28 and 28. The median ($M_e$) is the average of these two: $(28 + 28) / 2 = 28$.

The median is much less affected by outliers than the mean. If our oldest employee was 85 instead of 45, the sorted list would be [25, 28, 28, 30, 85], and the median would still be 28. This makes the median a better indicator of the "typical" value in datasets with skewed distributions or extreme values.

The Mode: The Most Frequent

The mode is simply the value that appears most often in the dataset.

Example: In our original age dataset [25, 30, 28, 45, 28], the value 28 appears twice, more than any other value. The mode is 28.

A dataset can have:

• One mode (unimodal): Like our example.
• Multiple modes (multimodal): E.g., [2, 3, 3, 4, 5, 5, 6] has modes 3 and 5.
• No mode: If all values appear with the same frequency (often, only once). E.g., [10, 20, 30, 40].

The mode is especially useful for categorical data (non-numerical data like "color" or "product type"), where mean and median don't make sense. For numerical data, it tells you the most common specific value.

Which Central Tendency Measure to Use?

• Use the mean for symmetric numerical data without significant outliers.
• Use the median for skewed numerical data or data with significant outliers.
• Use the mode for categorical data or when identifying the most common numerical value is important.

Often, reporting both the mean and median gives a more complete picture, especially if they differ significantly, suggesting skewness or outliers.

Histogram of the sample ages [25, 30, 28, 45, 28], showing the calculated mean, median, and mode. Notice how the single higher value (45) pulls the mean slightly higher than the median and mode.

Measures of Spread: Quantifying Variability

Knowing the center of your data is only part of the story. You also need to know how spread out the data points are. Are they all clustered tightly around the mean, or are they widely dispersed? Measures of spread, or variability, answer this question.

The Range

The range is the simplest measure of spread. It's the difference between the maximum and minimum values in the dataset.

Range = Maximum Value - Minimum Value

Example: For our ages [25, 28, 28, 30, 45]: Maximum = 45 Minimum = 25 Range = 45 - 25 = 20 years.

The range gives a quick sense of the total span of the data, but like the mean, it's highly sensitive to outliers. Just one very high or very low value dramatically affects the range. It also doesn't tell you anything about how the data is distributed between the extremes.

Variance ($s^2$)

Variance measures the average squared difference of each data point from the mean. It gives you a sense of the overall spread. A larger variance means data points tend to be further from the mean; a smaller variance means they tend to be closer.

The formula for sample variance ($s^2$) looks a bit complex, but the idea is straightforward:

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

Let's break it down:

1. $(x_i - \bar{x})$: Find the difference between each data point ($x_i$) and the mean (\bar{x). These are called deviations.
2. $(x_i - \bar{x})^2$: Square each deviation. This makes all values positive (so negative and positive deviations don't cancel out) and emphasizes larger deviations.
3. $\sum_{i=1}^{n} (x_i - \bar{x})^2$: Sum up all the squared deviations.
4. $/ (n-1)$: Divide by the number of observations minus one ($n-1$). We use $n-1$ for a sample variance because it gives a better, unbiased estimate of the true population variance. If you were calculating variance for the entire population, you would divide by $N$ (the population size). For introductory purposes, we usually work with samples.

Example: Using ages [25, 28, 28, 30, 45] and mean $\bar{x} = 31.2$:

• Deviations $(x_i - \bar{x})$:
  • $25 - 31.2 = -6.2$
  • $28 - 31.2 = -3.2$
  • $28 - 31.2 = -3.2$
  • $30 - 31.2 = -1.2$
  • $45 - 31.2 = 13.8$
• Squared Deviations $(x_i - \bar{x})^2$:
  • $(-6.2)^2 = 38.44$
  • $(-3.2)^2 = 10.24$
  • $(-3.2)^2 = 10.24$
  • $(-1.2)^2 = 1.44$
  • $(13.8)^2 = 190.44$
• Sum of Squared Deviations: $38.44 + 10.24 + 10.24 + 1.44 + 190.44 = 250.8$
• Sample Variance $s^2 = 250.8 / (5 - 1) = 250.8 / 4 = 62.7$

The variance is 62.7. What does this number mean? It's in "squared years," which isn't very intuitive. That's where standard deviation comes in.

Standard Deviation ($s$ or $\sigma$)

The standard deviation is simply the square root of the variance. It's often preferred because it brings the measure of spread back into the original units of the data.

Standard Deviation ($s$) = $\sqrt{\text{Variance}} = \sqrt{s^2}$

Example: For our ages, the variance $s^2 = 62.7$. The standard deviation $s = \sqrt{62.7} \approx 7.92$ years.

The standard deviation gives you a measure of the typical or average distance of the data points from the mean. A standard deviation of 7.92 years suggests that, on average, the employees' ages fall about 7.9 years away from the mean age of 31.2.

• Low Standard Deviation: Data points are clustered tightly around the mean.
• High Standard Deviation: Data points are spread out over a wider range of values.

Like the mean, variance and standard deviation are sensitive to outliers because they are based on the mean and involve squared deviations, which heavily weight extreme values.

These summary statistics (mean, median, mode, range, variance, standard deviation) are fundamental building blocks for understanding your data. Calculating them is often the very first step in any Exploratory Data Analysis, providing a concise quantitative description of the dataset's characteristics before you proceed to visualization or more complex modeling.