Comprehensive Study Guide: Descriptive Statistics, Probability, and Distributions

BASIC CONCEPTS OF STATISTICS

Statistics Defined: Statistics is the art of learning from data. It is a science that deals with the collection, processing, analysis, and interpretation of numerical data.
Population: Refers to the total collection of elements of interest. Examples include all cars produced by a specific company in the last year or all students enrolled in an institution. Populations can be finite or infinite in size.
Sample: A smaller subgroup of elements selected from the population for selective examination. Unlike populations, samples are always finite.
Variable: A characteristic that changes or varies over time or across different experimental units. - Example: Body temperature is a variable that varies with time for an individual and varies from individual to individual at a particular time.
Experimental Unit: The specific individual or object on which a variable is measured.
Data Categories: - Univariate Data: Obtained when a single variable is measured on a single unit (e.g., blood pressure of employees). - Bivariate Data: Obtained when two variables are measured on a single unit (e.g., blood pressure and weight). - Multivariate Data: Obtained when more than two characteristics are measured on a single unit.
Categories of Variables: - Qualitative Variable: Measures a quality or characteristic (e.g., skin color: fair, wheatish, black; performance: excellent, average, poor). - Quantitative Variable: Measures a quantity or numerical amount (e.g., weight, height, marks). - Discrete Variable: Can assume only finite or countably infinite values (e.g., number of students). - Continuous Variable: Can assume an infinite number of values between two specific limits. Between any two selected values, a third value can always be found (e.g., time, temperature, weight).

DATA REPRESENTATION

Observations and numerical findings must be presented systematically to grasp essential characteristics.
Frequency Tables: Used for data sets with a relatively small number of distinct values.
Line Graph: Data is plotted with values on the x-axis and corresponding frequencies on the y-axis.
Ungrouped vs. Grouped Distribution: - Ungrouped: Lists every distinct data value with its frequency. - Grouped: Condenses large data sets using class intervals (e.g., salary-group 40-44).
Histogram: Composed of rectangles over class intervals on a horizontal scale. - The areas of the rectangles are proportional to the class frequencies. - For equal class intervals, heights are proportional to frequencies. - For unequal intervals, heights are proportional to the ratio of frequencies to the class width.
Frequency Polygon: Obtained by joining points with straight lines where the abscissae are mid-points (class marks) and ordinates are frequency. The polygon is closed by joining the midpoints of the horizontal axis immediately preceding and succeeding the observed intervals.
Frequency Curve: A smooth, free-hand version of the frequency polygon, typically used when class intervals are of very small width.
Cumulative Frequency Curve (Ogive): - Less Than Type: Cumulative frequencies plotted against upper class limits. - Greater Than Type: Cumulative frequencies plotted against lower class limits. - The curves help illustrate the nature of the distribution.

DESCRIPTIVE MEASURES: CENTRAL TENDENCY

Measures of Central Tendency (Averages): Representatives of the entire distribution that describe central location.
Arithmetic Mean ( $\bar{x}$ ): - Defined as the sum of observations divided by the sample size. - Ungrouped Formula: $\bar{x} = \frac{1}{n} \sum_{i=1}^n x_i$ - Frequency Distribution Formula: $\bar{x} = \frac{1}{N} \sum_{i=1}^n f_i x_i$ where $N = \sum f_i$ - In grouped distributions, $x_i$ is the mid-point (class mark). - Shifting of Origin and Scale: To reduce arithmetic for large values, use $u = \frac{x-a}{h}$ (where $a$ is assumed mean and $h$ is scale). The mean is then $\bar{x} = a + h\bar{u}$ .
Median: The positional average that divides the data into two equal parts when arranged in order. - Odd $n$ : Value in position $\frac{n+1}{2}$ . - Even $n$ : Average of values in positions $\frac{n}{2}$ and $\frac{n}{2} + 1$ . - Grouped Formula: $\text{Median} = l + \frac{h}{f} \left( \frac{N}{2} - C \right)$ $(l = \text{lower limit of median class, } f = \text{median class frequency, } C = \text{cf of pre-median class})$ .
Mode: The value that occurs with the greatest frequency. - Grouped Formula: $\text{Mode} = l + \frac{h(f_m - f_1)}{2f_m - (f_1 + f_2)}$ $(f_m = \text{modal frequency, } f_1 = \text{pre-modal, } f_2 = \text{post-modal})$ .
Geometric Mean (G): The $N^{th}$ root of the product of $N$ observations. Generally expressed as: $\ln(G) = \frac{1}{N} \sum f_i \ln(x_i)$
Harmonic Mean (H): The reciprocal of the arithmetic mean of the reciprocals of the data. - Formula: $\frac{1}{H} = \frac{1}{N} \sum \frac{f_i}{x_i}$ - Useful for true averages of speed or population growth.
Partition Values: - Quartiles ( $Q_1, Q_2, Q_3$ ): Divide data into 4 equal parts. - Deciles ( $D_1…D_9$ ): Divide into 10 equal parts. - Percentiles ( $P_1…P_{99}$ ): Divide into 100 equal parts. - $Q_i$ Formula: $Q_i = l + \frac{h}{f} \left( \frac{iN}{4} - C \right)$

DESCRIPTIVE MEASURES: DISPERSION

Measures of dispersion describe the scatteredness or spread of data about an average.
Range: Difference between the largest and smallest observations.
Inter-quartile Range (IQR): $Q_3 - Q_1$ . It utilizes the middle 50% of the data.
Variance ( $\sigma^2$ ): Quantifies variability around the mean. Average of squared deviations. $\sigma^2 = \frac{1}{N} \sum f_i (x_i - \bar{x})^2$
Standard Deviation (S.D. or $\sigma$ ): Positive square root of variance. Also called root-mean-square deviation.
Computational Variants: $\sigma^2 = \frac{1}{N} \sum f_i x_i^2 - (\bar{x})^2$
Coefficient of Variation (C.V.): A dimensionless measure of relative variability. $\text{C.V.}_x = \frac{\sigma}{\bar{x}} \times 100$

MOMENTS AND DISTRIBUTION SHAPE

Moments: - Ordinary Moment ( $\mu'_r$ ): Measured about any point $x = a$ . $\mu'_r = \frac{1}{N} \sum f_i (x_i - a)^r$ - Central Moment ( $\mu_r$ ): Measured about the mean. $\mu_r = \frac{1}{N} \sum f_i (x_i - \bar{x})^r$ - Relations: $\mu_2 = \mu'_2 - (\mu'_1)^2$ ; $\mu_3 = \mu'_3 - 3\mu'_2\mu'_1 + 2(\mu'_1)^3$ .
Pearson’s Coefficients: $\beta_1 = \frac{\mu_3^2}{\mu_2^3}; \gamma_1 = \sqrt{\beta_1}$ $\beta_2 = \frac{\mu_4}{\mu_2^2}; \gamma_2 = \beta_2 - 3$
Skewness: Measure of asymmetry. - Positive (Right): Long tail to the right. Mean > Mode. - Negative (Left): Long tail to the left. Mode > Mean. - Karl Pearson’s Coefficient: $\text{Skewness} = \frac{\text{Mean} - \text{Mode}}{\text{S.D.}}$
Kurtosis: Measure of the flatness or peakedness of a distribution curve. - Mesokurtic: Normal curve ( $\beta_2 = 3$ ). - Leptokurtic: More peaked than normal ( $\beta_2 > 3$ ). - Platykurtic: Flatter than normal ( $\beta_2 < 3$ ).

PROBABILITY CONCEPTS

Random Experiment: Process where the outcome is not unique but belongs to a set of possible outcomes.
Definitions: - Simple Event: A single possible outcome. - Exhaustive Events: Total possible outcomes of an experiment. - Mutually Exclusive: Occurrence of one prevents the occurrence of all others. - Equally Likely: No reason to expect one over others.
Classical Probability: Based on a trial with $n$ exhaustive, ME, and equally likely cases. $P(A) = \frac{\text{number of favorable cases } m}{\text{number of exhaustive cases } n}$
Statistical (Empirical) Probability: Based on relative frequency in a long run of experiments. $P(A) = \lim_{n \to \infty} \frac{m}{n}$
Axiomatic Approach: Probability is a function satisfying: 1. $0 \le P(A) \le 1$ 2. $P(S) = 1$ 3. If $A$ and $B$ are ME, $P(A \cup B) = P(A) + P(B)$ .

PROBABILITY LAWS

Addition Law (Total Probability): $P(A \cup B) = P(A) + P(B) - P(A \cap B)$
Conditional Probability: Probability of $A$ given $B$ has occurred. $P(A|B) = \frac{P(A \cap B)}{P(B)}$
Independent Events: If $P(A|B) = P(A)$ , occurrence of one does not affect the other.
Multiplication Law: $P(A \cap B) = P(B)P(A|B) = P(A)P(B|A)$
Bayes' Rule: For a set of ME and exhaustive events $B_1, …, B_k$ : $P(B_i|A) = \frac{P(B_i)P(A|B_i)}{\sum_{j=1}^k P(B_j)P(A|B_j)}$ - $P(B_i)$ : A priori probabilities. - $P(B_i|A)$ : Posterior probabilities.

RANDOM VARIABLES AND DISTRIBUTIONS

Random Variable (X): A function assigning a real value to each outcome in a sample space.
Discrete Random Variable: Assumes finite or countably infinite values. - Probability Mass Function (PMF): $p(x) = P{X = x}$ , where $\sum p(x) = 1$ .
Continuous Random Variable: Can take any value in an interval. - Probability Density Function (PDF): $f(x)$ , where $\int_{-\infty}^{\infty} f(x)\,dx = 1$ .
Distribution Function (CDF): $F(x) = P{X \le x}$ . - For continuous: $f(x) = \frac{dF(x)}{dx}$ .
Mathematical Expectation (E(X)): - Discrete: $E(X) = \sum x_i p_i$ - Continuous: $E(X) = \int_{-\infty}^{\infty} x f(x)\,dx$ - Mean (\mu): Equals $E(X)$ . - Variance (\sigma^2): Equals $E(X - \mu)^2 = E(X^2) - [E(X)]^2$ .
Properties of Expectation: - $E(aX + b) = aE(X) + b$ - $E(X + Y) = E(X) + E(Y)$ - $E(XY) = E(X)E(Y)$ if independent.

CHEBYSHEV’S INEQUALITY

Provides a bound on probability when only mean and variance are known.
Theorem: For any random variable $X$ with mean $\mu$ , variance $\sigma^2$ , and for any constant $k > 0$ : $P {|X - \mu| \ge k} \le \frac{\sigma^2}{k^2}$
Often expressed as: $P {|X - \mu| < k\sigma} \ge 1 - \frac{1}{k^2}$

SPECIAL DISCRETE DISTRIBUTIONS

Binomial Distribution: Arises from $n$ independent Bernoulli trials (success/failure) with constant probability $p$ . - PMF: $P{X = x} = \binom{n}{x} p^x q^{n-x}$ - Mean: $np$ ; Variance: $npq$ .
Poisson Distribution: Limiting case of Binomial where $n \rightarrow \infty$ and $p \rightarrow 0$ , such that $np = \lambda$ (constant). - PMF: $P(x) = \frac{e^{-\lambda} \lambda^x}{x!}$ - Mean: $\lambda$ ; Variance: $\lambda$ .
Hypergeometric: Sampling without replacement from a finite population.
Geometric: Number of failures before the first success. - PMF: $P{X = x} = q^x p$
Multinomial: Extension of Binomial where trials have more than two outcomes.

SPECIAL CONTINUOUS DISTRIBUTIONS

Uniform Distribution: Constant probability over interval $[a, b]$ . - PDF: $f(x) = \frac{1}{b-a}$ .
Normal Distribution: Bell-shaped symmetric curve. - PDF: $f(x) = \frac{1}{\sigma\sqrt{2\pi}} e^{-\frac{(x-\mu)^2}{2\sigma^2}}$ - Standard Normal Variate (Z): $Z = \frac{x-\mu}{\sigma}$ . - Properties: Mean = Median = Mode; Area under curve is 1; 68.26% data in $\mu \pm \sigma$ , 99.73% in $\mu \pm 3\sigma$ .
Exponential Distribution: Model for time between events; possesses "lack of memory" property. - PDF: $f(x) = ae^{-ax}$ for $x \ge 0$ . - Mean: $\frac{1}{a}$ ; Variance: $\frac{1}{a^2}$ .
Gamma Distribution: Extensions of Exponential; used in queuing and reliability.
Beta Distribution: Defined over interval $[0, 1]$ ; used for proportions.

CURVE FITTING AND LEAST SQUARES

Method of Least Squares: Determines the "best-fit" curve by minimizing the sum of the squares of vertical deviations (errors).
Straight Line Fitting ( $y = a + bx$ ): - Normal Equations: 1. $\sum y_i = na + b\sum x_i$ 2. $\sum x_i y_i = a\sum x_i + b\sum x_i^2$
Polynomial of Degree k: Requires solving a system of $k+1$ linear equations.
Non-Polynomial Curves: Transformed into linear forms (e.g., $y = ae^{bx}$ becomes $\ln(y) = \ln(a) + bx$ ).

CORRELATION AND REGRESSION

Correlation: Measures the degree of linear association between two variables. - Scatter Diagram: Graph of $(x, y)$ points. Concentrated points indicate strong correlation. - Karl Pearson’s Coefficient ( $r$ ): $r_{xy} = \frac{\text{Cov}(x, y)}{\sigma_x \sigma_y}$ - Range: $-1 \le r \le +1$ . - $r = 0$ : Uncorrelated (no linear relationship).
Spearman’s Rank Correlation (\rho): Used for data arranged by merit/rank. $\rho = 1 - \frac{6 \sum d_i^2}{n(n^2 - 1)}$
Regression: Predicts the dependent variable based on an independent variable. - Line of Regression of Y on x: $y - \bar{y} = r \frac{\sigma_y}{\sigma_x} (x - \bar{x})$ - Regression Coefficients: $b_{yx}$ ; $r$ is the geometric mean of the two regression coefficients: $r = \pm \sqrt{b_{yx} b_{xy}}$ .
Angle between Regression Lines: If uncorrelated, lines are perpendicular. If perfectly correlated, lines coincide.

EXAMPLES AND QUANTITATIVE DATA

Example 24.15: Male employees. 21% smoke cigarettes, 5% cigar, 3% both. smokers = $0.21 + 0.05 - 0.03 = 0.23$ (23%). Non-smokers = 77%.
Example 24.25: Monograph acceptance. Odds 3:2, 4:3, 2:3. - $P(A) = 3/5, P(B) = 4/7, P(C) = 2/5$ . - Probability all favorable = $24/175$ .
Example 24.38: Assembly line items. Mean 50, Var 25. - $P(X > 75) \le 50/75 = 2/3$ . - $P(40 < X < 60) \ge 0.75$ .
Example 24.43: Bombing target. 50% hit chance. Needs 2 direct hits. To ensure 99% destruction chance, at least 11 bombs must be dropped.
Example 24.59: Chip defect. 20% defective in lot of 100. Mean 20, S.D. 4. - $P(\text{at most 15 defective}) = 0.1292$ .

QUESTIONS & DISCUSSION

Case Study: Coin Age (Q2): Measuring age of 40 coins (Age = Current year - Year on coin). Descriptive distribution was analyzed via histogram.
Transistor Gain (Q4): 90 transistors measured. Calculations performed for mean, median, and modal gain using grouped data.
Nicotine Content (Q6): 40 cigarettes measured. Mean, median, and specific standard deviation were extracted from values like 1.92, 1.09, etc.
A priori vs Posterior: Discussion in Bayes’ Rule context identifies initial estimates vs revised estimates after evidence (e.g., medication results or machine tests).