Descriptive Statistics: Central Tendency, Variation, and Shape

Overview: Central Tendency, Variation, and Shape

The data are described using three related ideas: central tendency (where the data center around), variation (how spread out the data are), and shape (the overall form of the distribution).
Normal curve as a reference: understanding center, spread, and shape helps us describe data and compare datasets.
Two practical examples from the transcript emphasize the same mean but different dispersion, illustrating why variation matters (means can be identical while spread differs).

Central Tendency

Central tendency answers: what value best represents the center of the data?
Arithmetic mean (the mean or average)
- Notation: $\bar{x}$ (read as “x-bar").
- Formula: $\bar{x} = rac{1}{n} ightsum{i=1}^{n} xi$
- Interpretation: provides the balance point of the data; sensitive to extreme values/outliers.
- Example from the transcript: two samples with five values each can have means 13 and 14, respectively, even if the distributions look different.
Median
- Definition: the middle value in a data set when ordered from smallest to largest.
- For odd n: the middle value; for even n: the average of the two middle values.
- Position: with n values, the median is the value at rank $\frac{n+1}{2}$ when ordered; specifically in even cases, average of the $\frac{n}{2}$ and $\frac{n}{2}+1$ values.
- Strengths/limitations: not sensitive to extreme values, but does not always reflect spread; useful for skewed data.
- Relation to shape: If the mean ≈ median, the distribution is likely symmetric; if mean > median, distribution tends to be right-skewed; if mean < median, left-skewed.
Mode
- Definition: the most frequent value (or category) in the data.
- Can be multiple modes (multimodal) or no mode (all values distinct).
- Characteristics: not influenced by extreme values; useful for categorical data as well as numerical data.
Practical note on reporting
- In skewed data, it is common to report both mean and median to convey center and shape.

Variation and Dispersion (Spread)

Why dispersion matters:
- The same mean can describe different datasets; dispersion reveals how far data are spread around the center.
- More dispersion ⇒ greater spread; tighter clustering ⇒ smaller dispersion.
Range
- Definition: the difference between the maximum and minimum values.
- Formula: $\text{Range} = x{(n)} - x{(1)}$
- Pros/Cons: simple but highly sensitive to outliers and does not reflect internal distribution.
Variance and Standard Deviation
- Intuition: quantify how far each value deviates from the mean on average (squared deviations for variance; back to units for standard deviation).
- Sample variance: $s^2 = \frac{1}{n-1} \sum{i=1}^{n} (xi - \bar{x})^2$
- Standard deviation: $s = \sqrt{ s^2 } = \sqrt{ \frac{1}{n-1} \sum{i=1}^{n} (xi - \bar{x})^2 }$
- Why square differences: to avoid cancellation of positive and negative deviations and to emphasize larger deviations.
- Relationship: Variance is the square of the standard deviation; standard deviation is the square root of the variance.
Coefficient of variation (CV)
- Definition: a relative measure of dispersion, expressed as a ratio of the standard deviation to the mean.
- Formula: $\text{CV} = \frac{s}{\bar{x}}$ (often written as a percentage by multiplying by 100%).
- Use: enables comparison of variability across datasets with different units or scales (e.g., GRE vs GMAT scores).
Z-scores (standardization)
- Definition: how many standard deviations a value is from the mean.
- Formula: $z = \frac{x - \bar{x}}{s}$
- Interpretation: z = 0 means the value equals the mean; positive z means above the mean; negative z means below the mean.
- Role in later topics: used in hypothesis testing and to compare values from different distributions or scales.
Steps for computing standard deviation (summary):
1) Compute the mean $\bar{x}$ .
2) Subtract the mean from each value, then square the result.
3) Sum these squared deviations.
4) Divide by $n-1$ (for a sample) to obtain the variance, then take the square root to obtain the standard deviation.
Practical interpretation:
- In comparing datasets with the same mean, the one with smaller standard deviation is more tightly clustered around the mean.
- In finance, a smaller CV or a smaller standard deviation relative to mean implies less risk/variance; a larger CV implies higher relative variability.

Shape and Distribution (Normal Curve, Skewness, Kurtosis)

Shape concept
- A distribution can be symmetric (roughly bell-shaped), right-skewed (longer tail to the right), or left-skewed (longer tail to the left).
- Skewness focuses on tail asymmetry; kurtosis on peakedness.
Skewness (asymmetry) and the mean/median relationship
- If the mean < median → left-skewed (tail on the left).
- If the mean > median → right-skewed (tail on the right).
- If mean ≈ median → roughly symmetric (normal-like).
Kurtosis (peakedness)
- Mesokurtic: standard, moderate peak (kurtosis similar to normal).
- Leptokurtic: peaky distribution (high peak; data concentrated near the center).
- Platykurtic: flat distribution (lower peak; more spread out).
Normal distribution as a reference
- In a symmetric, bell-shaped distribution, the empirical rule (68-95-99.7%) applies.
- Empirical rule: For a normal distribution with mean $\mu$ and standard deviation $\sigma$ ,
- Approximately $68\%$ of data lie within $\mu \pm \sigma$ .
- Approximately $95\%$ lie within $\mu \pm 2\sigma$ .
- Approximately $99.7\%$ lie within $\mu \pm 3\sigma$ .
- Values beyond ±3σ are considered outliers in a well-behaved normal distribution.
Skewness and distribution interpretation using quartiles/boxplots (brief mental checks)
- If the distribution is right-skewed, the right tail is longer; left-skewed has a longer left tail.
- In a boxplot: a longer right tail indicates right skew; longer left tail indicates left skew.

Quartiles, Five-Number Summary, and Box Plots

Quartiles
- Q1 (first quartile): 25th percentile
- Q2 (second quartile / median): 50th percentile
- Q3 (third quartile): 75th percentile
- Position formulas (rank-based; interpolation for non-integer ranks):
- Q1 position: $\frac{n+1}{4}$
- Q2 position: $\frac{n+1}{2}$
- Q3 position: $\frac{3(n+1)}{4}$
- Interpolation when the rank is not an integer (e.g., 12.5): take the average of the two surrounding values.
Interquartile Range (IQR)
- Definition: distance between Q3 and Q1.
- Formula: $\text{IQR} = Q3 - Q1$
Five-number summary
- The set: min, Q1, median (Q2), Q3, max.
Box plots (box-and-whisker plots)
- Visual display of the five-number summary: gives a sense of center, spread, and skewness.
- Box center line shows the median; whiskers show spread beyond the quartiles.
Left vs. right skew via quartiles
- If Q1 − min > max − Q3, boxplot indicates left-skew; if the opposite, right-skew (beam toward the longer tail).
How to read skewness from the data description
- If the distribution is symmetric, the center of the box aligns with the centerline; if skewed, the median shifts toward the shorter side and whiskers differ in length.

Populations, Samples, and Notation

Descriptive vs inferential statistics
- Descriptive statistics describe a sample (or population when data exist) using sample measures.
- Population parameters describe population quantities; sample statistics estimate them.
Notation distinctions
- Population: mean $\mu$ ; variance $\sigma^2$ ; standard deviation $\sigma$ .
- Sample: mean $\bar{x}$ ; variance $s^2$ ; standard deviation $s$ .
- Relationship: $\mu\approx\bar{x}$ , $\sigma^2\approx s^2$ when the sample reflects the population well.
Example notations
- If the population size is N, the population mean is $\mu = \frac{1}{N} \sum{i=1}^{N} Xi$ .
- The population variance is $\sigma^2 = \frac{1}{N} \sum{i=1}^{N} (Xi - \mu)^2$ .
Important practical note
- In practice, most statistics are computed from samples and used to infer about populations; parameters are fixed but often unknown.

Covariance and Correlation (Two-Variable Relationships)

Covariance
- Measures the direction of a linear relationship (positive or negative) between two variables; does not by itself indicate strength clearly.
- Formula (sample): $\mathrm{Cov}(X,Y) = \frac{1}{n-1} \sum{i=1}^{n} (Xi - \bar{X})(Y_i - \bar{Y})$
- Sign indicates direction, magnitude indicates co-movement; values depend on the units of X and Y.
Correlation (Pearson r)
- Normalized measure of linear relationship; unit-free; ranges from -1 to 1.
- Formula (sample): $r = \frac{\mathrm{Cov}(X,Y)}{sX sY}$
- Population counterpart: $\rho = \frac{\mathrm{Cov}(X,Y)}{\sigmaX \sigmaY}$
- Interpretation:
- r close to 1: strong positive linear relationship
- r close to -1: strong negative linear relationship
- r near 0: little to no linear relationship
- Important caveat: correlation does not imply causation; it only describes association.
Quick takeaway from scatter plots
- Covariance/Correlation summarize the strength and direction of a linear association across paired observations.

Practical Examples and Interpretations from the Transcript

Income example with two localities (same mean, different spread)
- Locality A: 20 households, incomes from 70k to 150k, mean ≈ 100k, moderate dispersion.
- Locality B: 20 households, mean ≈ 100k, but one extreme value (e.g., milk income) creates more dispersion, illustrating how mean alone can be misleading without variation.
Real-world uses of central tendency and dispersion
- GDP example: mean per-capita income vs. distribution of income (the mean can mask inequality).
- Median vs. mean in reporting house prices or other skewed data; median often preferred when outliers exist.
Z-scores example (SAT-like scores)
- Given a dataset with mean 490 and standard deviation 100, a score of 620 has z-score $z = \frac{620-490}{100} = 1.3$ , i.e., between the first and second standard deviation above the mean.
Population vs. sample and notation in context
- Emphasizes that many statistics are computed from samples and used to infer properties of populations; parameters describe populations; statistics describe samples.
One more practical point: the Empirical Rule (68-95-99.7%) applies to roughly normal data and helps identify outliers and assess whether a dataset is approximately normal.

Quick Reference: Key Formulas and Concepts (LaTeX)

Mean (sample): $\bar{x} = \frac{1}{n} \sum{i=1}^{n} xi$
Range: $\text{Range} = x{(n)} - x{(1)}$
Variance (sample): $s^2 = \frac{1}{n-1} \sum{i=1}^{n} (xi - \bar{x})^2$
Standard deviation (sample): $s = \sqrt{s^2} = \sqrt{ \frac{1}{n-1} \sum{i=1}^{n} (xi - \bar{x})^2 }$
Coefficient of variation: $\text{CV} = \frac{s}{\bar{x}}\times 100\%$
Z-score: $z = \frac{x - \bar{x}}{s}$
Population mean: $\mu = \frac{1}{N} \sum{i=1}^{N} Xi$
Population variance: $\sigma^2 = \frac{1}{N} \sum{i=1}^{N} (Xi - \mu)^2$
Covariance (sample): $\mathrm{Cov}(X,Y) = \frac{1}{n-1} \sum{i=1}^{n} (Xi - \bar{X})(Y_i - \bar{Y})$
Correlation (sample): $r = \frac{\mathrm{Cov}(X,Y)}{sX sY}$
Empirical rule (normal distribution):
- $P(|X-\mu| \leq \sigma) \approx 0.68$
- $P(|X-\mu| \leq 2\sigma) \approx 0.95$
- $P(|X-\mu| \leq 3\sigma) \approx 0.997$
Quartiles (rank-based; interpolation for non-integers)
- Q1 position: $\frac{n+1}{4}$
- Q2 (median) position: $\frac{n+1}{2}$
- Q3 position: $\frac{3(n+1)}{4}$
Interquartile Range: $\text{IQR} = Q3 - Q1$
Five-number summary: min, Q1, median, Q3, max
Box plots: visual representation of the five-number summary and the skewness of the distribution

Notes: The content above mirrors the key ideas and examples discussed in the transcript, organized into comprehensive study notes. It emphasizes how central tendency and variation together shape our understanding of data, how to interpret skewness and kurtosis, and how to use standard tools like z-scores, CV, and quartiles to compare and describe datasets. If you want, I can also tailor this into a compact cheat sheet or expand any section with additional worked examples.