STATISTICS COURSE
Comprehensive Explanation Notes
A Complete Guide to Statistical Analysis and Inference
Topics Covered:
• Descriptive Statistics
• Probability and Probability Distributions
• Statistical Inferences from Sampling
• Hypothesis Testing and Non-parametric Methods
TABLE OF CONTENTS
1. Descriptive Statistics...................................................................3
2. Probability and Probability Distributions...................................................................4
3. Statistical Inferences from Sampling...................................................................5-6
4. Hypothesis Testing...................................................................6-7
1. DESCRIPTIVE STATISTICS
1.1 Introduction to Descriptive Statistics
Descriptive statistics involves methods for organizing, summarizing, and displaying data. Unlike
inferential statistics which makes predictions about populations, descriptive statistics focuses on
describing the characteristics of a given dataset. The main goal is to transform raw data into meaningful
information that can be easily understood and communicated.
1.2 Frequency Distribution
Definition: A frequency distribution is a table that shows the number of observations (frequency) that
fall into each of several categories or class intervals. It helps identify patterns and trends in data at a
glance.
Steps to Create a Frequency Distribution:
1. Determine the number of classes (usually between 5-20)
2. Calculate the class width: (Maximum - Minimum) ÷ Number of classes
3. Set up class intervals starting from the minimum value
4. Count observations in each class
5. Calculate relative frequencies (frequency ÷ total observations)
1.3 Measures of Central Tendency
Central tendency measures describe the center or typical value of a dataset. The three main measures
are:
Measure Definition Formula When to Use
Mean Average of all values Σx ÷ n For normally distributed data without extreme outliers
Median Middle value when data is
Middle
ordered
value of ordered
For data
skewed distributions or when outliers are present
Mode Most frequently occurringValue
valuewith highest frequency
For categorical data or to identify most common value
1.4 Measures of Dispersion (Variability)
Dispersion measures indicate how spread out the data is from the center. Key measures include:
Range: The difference between the maximum and minimum values. Simple but affected by outliers.
Variance (σ2): Average of squared deviations from the mean. Measures average squared distance
from center.
Standard Deviation (σ): Square root of variance. Most commonly used measure, in same units as
original data.
Coefficient of Variation (CV): (Standard Deviation ÷ Mean) × 100. Allows comparison across datasets
with different units.
1.5 Shapes of Frequency Distribution
Skewness: Measures asymmetry of distribution. Positive skew: tail on right (mean > median). Negative
skew: tail on left (mean < median). Zero skew: symmetric (mean = median = mode).
Kurtosis: Measures the 'peakedness' of distribution. Leptokurtic: sharp peak. Mesokurtic: normal.
Platykurtic: flat distribution.
2. PROBABILITY AND PROBABILITY DISTRIBUTIONS
2.1 Fundamentals of Probability
Probability is the likelihood or chance that an event will occur. It ranges from 0 (impossible) to 1
(certain). P(Event) = Number of favorable outcomes ÷ Total number of possible outcomes.
2.2 Types of Events and Probability Rules
Single Events: Probability of one specific outcome occurring.
Multiple Events: Probability involving two or more events.
Addition Rule: P(A or B) = P(A) + P(B) - P(A and B)
Multiplication Rule: P(A and B) = P(A) × P(B|A) where P(B|A) is conditional probability.
Complement Rule: P(A') = 1 - P(A)
2.3 Discrete Probability Distributions
Discrete distributions apply to countable outcomes (0, 1, 2, 3, ...). Key distributions include:
Binomial Distribution: Models number of successes in n independent trials with constant probability
p. Used for yes/no, pass/fail scenarios. Parameters: n (number of trials), p (probability of success).
Mean = np, Variance = np(1-p)
Poisson Distribution: Models number of events in fixed time/space interval when events occur at
constant rate. Used for rare events (accidents, defects, arrivals). Parameter: λ (lambda, rate
parameter).
Mean = λ, Variance = λ
2.4 Continuous Probability Distributions
Normal Distribution (Gaussian): Most important continuous distribution, bell-shaped and symmetric.
Defined by mean (µ) and standard deviation (σ). Many natural phenomena follow normal distribution.
Approximately 68% of data within 1σ, 95% within 2σ, 99.7% within 3σ (68-95-99.7 Rule).
Standard Normal Distribution: Special case with µ = 0 and σ = 1. Any normal distribution can be
converted using Z-score: Z = (X - µ) ÷ σ. Z-scores allow comparison of different measurements.
3. STATISTICAL INFERENCES FROM SAMPLING
3.1 Sampling and Sampling Distributions
Population vs. Sample: Population is the entire group of interest; sample is a subset selected for
study. Statistical inference uses sample data to make conclusions about the population.
Sampling Distribution: Distribution of a statistic (like sample mean) calculated from repeated random
samples. Key property: Sampling distribution of mean is approximately normal even if population isn't,
due to Central Limit Theorem.
3.2 Central Limit Theorem (CLT)
The Central Limit Theorem states that regardless of the shape of the population distribution, the
sampling distribution of the sample mean approaches a normal distribution as sample size increases.
This is true for sample sizes n ≥ 30 (or smaller if population is already normal). Mean of sampling
distribution = µ (population mean). Standard error (SE) = σ ÷ √n, where σ is population standard
deviation.
3.3 Estimation: Point and Interval Estimates
Point Estimate: Single value used to estimate population parameter (e.g., sample mean estimates
population mean). Provides no information about accuracy.
Interval Estimate (Confidence Interval): Range of values that likely contains the population
parameter. Provides margin of error and confidence level (typically 90%, 95%, or 99%).
3.4 Confidence Intervals
For Population Mean (σ known):
CI = x■ ± Zα/2 × (σ ÷ √n)
where x■ is sample mean, Z is critical value, α is significance level.
For Population Mean (σ unknown, using t-distribution):
CI = x■ ± tα/2 × (s ÷ √n)
where s is sample standard deviation, t uses (n-1) degrees of freedom. Used when population standard
deviation unknown (most real situations).
4. HYPOTHESIS TESTING
4.1 Fundamentals of Hypothesis Testing
Hypothesis testing is a statistical method to make decisions about population parameters based on
sample data. It involves setting up two competing hypotheses and using sample evidence to decide
which is more credible.
4.2 Types of Hypotheses
Null Hypothesis (H0): States no effect, no difference, or no relationship. Assumed true unless
evidence proves otherwise.
Example: H0: µ = 100
Alternative Hypothesis (H1 or Ha): What we're trying to prove. Can be one-tailed (directional) or
two-tailed (non-directional).
Example: H1: µ ≠ 100 (two-tailed) or H1: µ > 100 (one-tailed)
4.3 Type I and Type II Errors
H<sub>0</sub> is True H<sub>0</sub> is False
Reject H<sub>0</sub> Type I Error (α) Correct Decision
Fail to Reject H<sub>0</sub> Correct Decision Type II Error (β)
Type I Error (α): Rejecting true null hypothesis. Significance level usually set at 0.05 (5%).
Type II Error (β): Failing to reject false null hypothesis. Power = 1 - β is the probability of correctly
rejecting false H0.
4.4 Test Statistics and P-values
Test Statistic: Calculated from sample data to determine if null hypothesis should be rejected. Formula
depends on the type of test (t-test, z-test, etc.).
P-value: Probability of observing test statistic as extreme as or more extreme than observed, assuming
H0 is true. If p-value < α, reject H0. Smaller p-value indicates stronger evidence against H0.
4.5 Tests About Population Mean
One-Sample t-test: Tests if population mean equals hypothesized value. Test statistic: t = (x■ - µ0) ÷
(s ÷ √n), with (n-1) df.
Two-Sample t-test: Compares means of two independent populations. Assumes equal variances (or
uses Welch's correction). Test statistic: t = (x■1 - x■2) ÷ SE(difference)
Paired t-test: Compares means when data are dependent (matched pairs or before-after). Test
statistic: t = (d■) ÷ (sd ÷ √n) where d are differences.
4.6 Tests About Population Proportion
One-Sample Proportion Test: Tests if population proportion equals hypothesized value. Uses z-test
when np and n(1-p) both ≥ 5. Test statistic: z = (p■ - p0) ÷ √[p0(1-p0)/n]
Two-Sample Proportion Test: Compares proportions from two populations. Uses pooled proportion
for standard error calculation.
4.7 Goodness of Fit Test
Chi-Square Goodness of Fit Test: Tests if observed frequencies match expected frequencies under a
hypothesis. Compares categorical data to expected distribution. Test statistic: χ2 = Σ[(O - E)2 ÷ E],
where O = observed, E = expected frequency. Follows chi-square distribution with (k-1) degrees of
freedom, where k = number of categories.
4.8 Analysis of Variance (ANOVA)
One-Way ANOVA: Tests if means of three or more independent groups are equal. H0: µ1 = µ2 = ... =
µk. Partitions total variation into between-group and within-group variation. Test statistic: F = (MS
between) ÷ (MS within), where MS = mean square. Larger F indicates stronger evidence against H0.
Requires post-hoc tests (like Tukey) if H0 rejected.
4.9 Chi-Square Test of Independence
Purpose: Tests if two categorical variables are independent (related or not). Uses contingency table
showing frequency distribution of two variables. Test statistic: χ2 = Σ[(O - E)2 ÷ E]. Degrees of freedom
= (rows - 1) × (columns - 1). Large χ2 indicates relationship between variables.
4.10 Non-parametric Testing
Non-parametric tests don't assume specific probability distributions and work with ranks rather than raw
values. Use when: data are ordinal, distribution is non-normal, sample size is small, or outliers are
present.
Mann-Whitney U Test: Non-parametric alternative to two-sample t-test. Compares medians of two
independent groups using ranks.
Wilcoxon Signed-Rank Test: Non-parametric alternative to paired t-test. Tests if distribution of
differences is symmetric around zero.
Kruskal-Wallis Test: Non-parametric alternative to one-way ANOVA. Compares medians of three or
more independent groups using ranks.
==============================================================================
==
Key Takeaway: Statistics provides tools to describe data, understand probability, estimate parameters
from samples, and make informed decisions through hypothesis testing. Choose the right method
based on data type, distribution, sample size, and research question.