OpenIntro Statistics PDF Download – Diez, Barr & Cetinkaya-Rundel (4th Edition) BS Statistics

BS Statistics and BSCS students taking an Introductory Statistics course can download OpenIntro Statistics, Fourth Edition, by David Diez, Mine Çetinkaya-Rundel, and Christopher D Barr, free from the authors’ own OpenIntro project under a Creative Commons licence. It is a calculus-free introduction that starts from how data is collected and summarized, builds up probability and the major distributions, and then covers confidence intervals, hypothesis testing, and linear regression — the full first-course sequence most Pakistani university syllabi follow.

This book is aimed at students preparing for board and semester exams in Introductory Statistics, where both computation and correct interpretation are tested. Its 9 chapters move from data basics through inference and regression using real data sets — a stent-and-stroke clinical trial, a malaria vaccine study, Lending Club loan data — so each method is learned in the context of an actual research question, which makes both first-pass learning and closer-to-exam revision easier.

Book Overview

Chapter List

Chapter 1: Introduction to Data

Difficulty: Easy · Key topics: data matrix, case/observational unit, numerical vs. categorical variables, discrete vs. continuous, ordinal vs. nominal, explanatory and response variables, association vs. causation, observational study vs. experiment, population vs. sample, anecdotal evidence, bias

This opening chapter uses a real clinical trial — 451 at-risk patients randomly assigned to a stent-plus-medication group or a medication-only group — to show what a statistical investigation actually looks like, before any formulas appear. From that case study it builds the vocabulary every later chapter assumes: rows (cases) and columns (variables) in a data matrix, numerical variables that are either discrete or continuous, and categorical variables that are either ordinal or nominal. It then separates two ways data gets collected (an observational study that can only show association, versus a randomized experiment that can support a causal claim) and two ways a sample can go wrong (haphazard anecdotal evidence, and systematic bias from a non-random sampling method).

Key Points:

• A pair of variables is either associated or independent — never both at once
• Association does not imply causation; causation can only be inferred from a randomized experiment
• In a randomized experiment, one group typically receives a placebo so the effect of the treatment itself can be isolated
• A sample is a subset of a target population; conclusions from a biased sample do not generalize to the full population
• A numerical variable is discrete if it can only take specific values (often counts), and continuous if it can take any value in a range
• A categorical variable with a natural order (e.g. low/medium/high) is ordinal; one without an order (e.g. eye color) is nominal

Memory Tip: Before moving on, be able to say in one sentence why the stent study’s surprising result (8% more strokes in the treatment group) could not, on its own, prove stents are harmful — because an observed difference between two groups can always be partly due to chance, and the rest of the book exists to measure how much.

Common Mistake: Treating any difference seen between two groups in an observational study as proof that one variable caused the other, when only a randomized experiment supports a causal conclusion.

Important Questions:

• What is the difference between an observational study and a randomized experiment? An observational study only records existing data without assigning treatments, so it can show association at best; a randomized experiment assigns subjects to groups by chance, which is what allows a causal conclusion.
• Why does a biased sample fail to represent the population it was drawn from? Bias is a systematic (not random) tendency for a sampling method to over- or under-represent certain parts of the population, so increasing the sample size does not fix it — a larger biased sample is still biased.

Chapter 2: Summarizing Data

Difficulty: Easy · Key topics: mean, median, variance, standard deviation, interquartile range (IQR), outliers, skew and long tails, boxplots, categorical data summaries

This chapter is the toolkit for describing a data set in numbers and pictures, and the book is explicit that mastering it is essential for everything that follows. For numerical data it covers the mean as a balancing point, the median as the middle value when data are sorted (robust to outliers in a way the mean is not), and three measures of spread: variance, standard deviation, and the interquartile range used to draw a boxplot. It shows how a long tail on one side of a distribution identifies skew, and why outliers deserve investigation rather than automatic deletion. For categorical data it covers frequency and relative frequency tables, bar plots, and a short case study on a malaria vaccine trial that ties the numerical and categorical tools together in one real analysis.

Key Points:

• The sample mean is x̄ = (x1 + x2 + ... + xn) / n — the sum of observed values divided by the count
• The median is the middle value when data are sorted (the average of the two middle values if n is even), and is resistant to outliers
• The IQR is the length of the box in a boxplot: IQR = Q3 − Q1, the distance between the 25th and 75th percentiles
• Variance is the average squared distance from the mean; standard deviation is its square root and is in the same units as the data
• A distribution with a long left tail is left-skewed; a long right tail makes it right-skewed
• An outlier is an observation that is extreme relative to the rest of the data, and checking for outliers can reveal skew or a data-entry error

Practice Tip: Before trusting a mean, sketch the shape of the data first — on a strongly skewed distribution (like loan interest rates or income), the median describes a “typical” value far better than the mean, which gets pulled toward the long tail.

Common Mistake: Reporting only the mean for a skewed data set, which can be badly misleading when a handful of extreme values pull it away from where most of the data actually sit.

Important Questions:

• Why is the median preferred over the mean for a strongly skewed distribution? The mean is pulled in the direction of the long tail by extreme values, while the median — the middle value when data are sorted — is not affected by how extreme the tail values are, only by their rank.
• How is the interquartile range (IQR) calculated, and what does it represent? IQR = Q3 − Q1, the difference between the 75th and 25th percentiles; it represents the spread of the middle 50% of the data and is the length of the box in a boxplot.

Chapter 3: Probability

Difficulty: Medium · Key topics: Law of Large Numbers, disjoint outcomes and the Addition Rule, General Addition Rule, complement, independence and the Multiplication Rule, marginal vs. joint probability, conditional probability, Bayes’ Theorem, random variables, expected value, linear combinations of random variables

The book states plainly at the start of this chapter that mastering it is not required to apply the applied methods in the rest of the book — it is a theoretical foundation offered for a deeper understanding, and students on a tight syllabus can treat several of its later sections as optional. It defines probability as the long-run proportion of times an outcome occurs (the Law of Large Numbers), then builds the rules for combining probabilities: the Addition Rule for disjoint outcomes, the General Addition Rule for any two events, and the complement rule. It separates independent events (where the Multiplication Rule applies directly) from dependent ones, introduces conditional probability and Bayes’ Theorem for inverting a conditional probability, and closes by defining a random variable and its expected value, variance, and how both behave under a linear combination of two random variables.

Key Points:

• Addition Rule (disjoint outcomes): P(A or B) = P(A) + P(B); the General Addition Rule for any two events subtracts the overlap: P(A or B) = P(A) + P(B) − P(A and B)
• Complement: P(A) + P(A^c) = 1, so P(A) = 1 − P(A^c)
• Multiplication Rule for independent events: P(A and B) = P(A) × P(B) — valid only when A and B come from independent processes
• Conditional probability: P(A | B) = P(A and B) / P(B)
• Bayes’ Theorem inverts a conditional probability using the General Multiplication Rule and the sum of conditional probabilities across all outcomes
• Expected value of a discrete random variable: E(X) = Σ xi × P(X = xi), often written μ
• For independent random variables, Var(aX + bY) = a²Var(X) + b²Var(Y) — variances add, but only under independence

Memory Tip: Multiplying probabilities (P(A) × P(B)) is only valid for independent events; the moment two events might influence each other, switch to the General Multiplication Rule, P(A and B) = P(A|B) × P(B), which is always valid.

Common Mistake: Multiplying two probabilities together to get a joint probability without first confirming the two events are independent — this silently gives a wrong answer whenever the events are actually related.

Important Questions:

• What is the difference between the Addition Rule for disjoint outcomes and the General Addition Rule? The simple Addition Rule, P(A or B) = P(A) + P(B), only works when A and B cannot both happen; the General Addition Rule, P(A or B) = P(A) + P(B) − P(A and B), subtracts the double-counted overlap and works for any two events.
• When can you multiply two probabilities directly to find P(A and B)? Only when A and B are independent events from separate processes; otherwise the correct formula is the General Multiplication Rule, P(A and B) = P(A|B) × P(B).

Chapter 4: Distributions of Random Variables

Difficulty: Medium · Key topics: normal distribution, Z-score, Bernoulli trials, binomial distribution and its four conditions, normal approximation to the binomial, negative binomial distribution, geometric distribution, Poisson distribution

This chapter introduces the specific probability distributions the rest of the book leans on, starting with the normal distribution — the symmetric, bell-shaped curve the book calls “overwhelmingly the most common” — and the Z-score, which measures how many standard deviations an observation sits from the mean. It then covers discrete distributions built from repeated Bernoulli trials: the geometric distribution (trials until the first success), the binomial distribution (successes in a fixed number of trials, with a normal approximation once np and n(1−p) are both at least 10), and the negative binomial distribution (trials until the k-th success). It closes with the Poisson distribution for counting rare events over a fixed interval, governed by a single rate parameter, λ.

Key Points:

• Z-score: Z = (x − μ) / σ — the number of standard deviations an observation falls above or below the mean
• A binomial setting requires four conditions: independent trials, a fixed number of trials n, each trial classified success/failure, and a constant success probability p
• Binomial probability of exactly k successes in n trials: C(n,k) p^k (1−p)^(n−k), with mean μ = np
• The binomial is well-approximated by a normal distribution once np ≥ 10 and n(1−p) ≥ 10
• Geometric distribution gives the probability the first success occurs on trial n: (1−p)^(n−1) p
• Poisson distribution for counting events at rate λ: P(k events) = λ^k e^(−λ) / k!

Memory Tip: Always draw and shade the normal curve before computing a Z-score — the book repeats this as a rule, because a picture catches a sign error (left tail vs. right tail) that the formula alone will not.

Common Mistake: Forgetting that most statistical software reports the area to the left of a Z-score by default, then using that left-tail probability directly when the question asked for the area to the right.

Important Questions:

• What four conditions must hold for a binomial model to apply? The trials must be independent, the number of trials n must be fixed in advance, each trial must be classifiable as success or failure, and the probability of success p must be the same on every trial.
• How is a Z-score calculated, and what does it tell you? Z = (x − μ) / σ; it tells you how many standard deviations the value x falls above (positive) or below (negative) the distribution’s mean.

Chapter 5: Foundations for Inference

Difficulty: Medium · Key topics: point estimate, sampling variability, sampling error vs. bias, 95% confidence interval, margin of error, null and alternative hypotheses, p-value, significance level

This is the chapter where the book’s whole approach to inference is established, using the sample proportion as the running example before the later chapters generalize it to means and categorical data. It distinguishes sampling error (natural variation from one sample to the next) from bias (a systematic flaw in how the sample was collected, which a bigger sample never fixes), then shows how the Central Limit Theorem lets a sample proportion be modeled as nearly normal once independence and the success-failure condition hold. From there it builds the 95% confidence interval as point estimate ± 1.96 × SE, generalizes to any confidence level with a z* multiplier, and introduces the hypothesis-testing framework: a skeptical null hypothesis, an alternative, a p-value, and the 0.05 significance-level convention the book spends a sidebar explaining rather than assuming.

Key Points:

• Sampling error is natural variation between samples; bias is a systematic over- or under-estimate that a larger sample size does not correct
• 95% confidence interval for a nearly-normal point estimate: point estimate ± 1.96 × SE; more generally, point estimate ± z* × SE
• The margin of error is the quantity z* × SE in a confidence interval
• The null hypothesis (H0) is the skeptical default; the alternative hypothesis (HA) is the claim under consideration
• The p-value is the probability of seeing data at least as favorable to HA as the observed data, assuming H0 is true
• If the p-value is less than the significance level α, reject H0; otherwise, fail to reject H0 (this is never the same as proving H0 true)

Memory Tip: “Fail to reject” is a deliberate double negative, not sloppy wording — it signals that the data were not surprising enough to abandon the null hypothesis, which is different from the data proving the null hypothesis correct.

Common Mistake: Interpreting a 95% confidence interval as “there’s a 95% chance the true parameter is in this interval,” when it actually means that 95% of intervals built this way, across repeated sampling, would capture the true parameter.

Important Questions:

• What is the difference between sampling error and bias? Sampling error is natural, unavoidable variability between samples of the same size from the same population; bias is a systematic tendency to over- or under-estimate the true value, caused by a flawed sampling method, and does not shrink as sample size grows.
• What does it mean when a hypothesis test’s p-value is less than the significance level α? It means the observed data would be unusually unlikely if the null hypothesis were true, so the null hypothesis is rejected in favor of the alternative.

Chapter 6: Inference for Categorical Data

Difficulty: Medium · Key topics: single-proportion inference, success-failure condition, difference of two proportions, pooled proportion, chi-square goodness-of-fit test, chi-square test of independence, degrees of freedom for a two-way table

This chapter applies the Chapter 5 framework to several categorical-data settings. It reviews confidence intervals and hypothesis tests for one proportion, checking independence and the success-failure condition (at least 10 expected successes and 10 expected failures). It extends the same ideas to the difference of two proportions, using a pooled proportion estimate specifically when the null hypothesis claims the two proportions are equal. For data in tables rather than single proportions, it introduces the chi-square distribution: a goodness-of-fit test for whether observed counts in one categorical variable match an expected distribution, and a test of independence for whether two categorical variables in a two-way table are related, each requiring every expected cell count to be at least 5.

Key Points:

• The sampling distribution of p̂ is nearly normal when observations are independent and both np ≥ 10 and n(1−p) ≥ 10 (the success-failure condition)
• When H0: p1 − p2 = 0, use the pooled proportion, p̂_pooled = (successes in both groups) / (total cases in both groups), to check conditions and build the standard error
• Chi-square goodness-of-fit test statistic: χ² = Σ (O − E)² / E, valid when every expected count is at least 5
• Expected count in a two-way table cell: (row total × column total) / table total
• Degrees of freedom for a two-way chi-square test: df = (R−1) × (C−1), where R and C are the number of rows and columns
• A one-way table describes counts for one variable’s outcomes; a two-way table describes counts for combinations of two variables’ outcomes

Practice Tip: Before running any chi-square test, compute every expected count first and check that each is at least 5 — this single check (not the chi-square formula itself) is what most often invalidates a result in practice.

Common Mistake: Using the simple one-proportion standard error formula (with p̂) when testing a two-proportion hypothesis of equality, instead of switching to the pooled proportion, which the null hypothesis of equal proportions specifically calls for.

Important Questions:

• When should the pooled proportion be used in a two-proportion hypothesis test? Specifically when the null hypothesis states the two population proportions are equal (H0: p1 − p2 = 0); the pooled proportion combines both samples’ successes and totals to check conditions and compute the standard error under that assumption.
• What condition must be checked before trusting a chi-square test’s result? Every expected cell count in the table must be at least 5; if any expected count falls below 5, the chi-square approximation is not reliable.

Chapter 7: Inference for Numerical Data

Difficulty: Hard · Key topics: t-distribution, degrees of freedom, one-sample mean inference, paired data, difference of two means, power calculations, ANOVA, F-statistic, Bonferroni correction

This chapter moves from proportions to means, and introduces the t-distribution as the tool that replaces the normal distribution once the population standard deviation must itself be estimated from the sample. It covers one-sample inference for a mean using df = n − 1 degrees of freedom, how to treat paired data (differences within pairs, not two separate samples), and the difference of two independent means. A section on power calculations asks the reverse question — how large a sample is needed to reliably detect a real difference — before the chapter ends with ANOVA for comparing three or more group means at once using the F-statistic, plus the Bonferroni correction for the inflated error rate that comes from testing many pairs of groups afterward.

Key Points:

• Central Limit Theorem for a mean: x̄ is nearly normal with SE = σ / √n for a sufficiently large, independent sample
• For inference on a single mean, use the t-distribution with df = n − 1, since σ is estimated by the sample standard deviation s
• Rule of thumb: if n < 30 with no clear outliers, the data are typically assumed nearly normal; checking normality gets harder to justify from the data alone as n shrinks further
• Paired data are analyzed through the differences within each pair, reducing a two-sample problem to a one-sample problem on those differences
• ANOVA tests whether the mean outcome differs across two or more groups using an F-statistic; rejecting H0 does not by itself identify which groups differ
• Bonferroni correction for multiple comparisons: use a stricter significance level, α* = α / K, where K is the number of pairwise comparisons being made

Memory Tip: If ANOVA rejects H0 but a follow-up pairwise comparison finds no significant pair, that is not a contradiction — it only means the test could detect that some group differs somewhere, without having the power to pin down exactly which pair.

Common Mistake: Running many pairwise t-tests after ANOVA without adjusting the significance level, which inflates the chance of a false positive across all those comparisons — exactly the problem the Bonferroni correction exists to fix.

Important Questions:

• Why does inference for a single mean use the t-distribution instead of the normal distribution? Because the population standard deviation σ is unknown and must be estimated by the sample standard deviation s, which adds extra uncertainty that the t-distribution (with df = n − 1) accounts for through its heavier tails.
• What does the Bonferroni correction adjust, and why? It lowers the significance level used for each individual comparison to α* = α / K (K = number of comparisons), because testing many pairs of groups at the original α level would otherwise inflate the overall chance of a false positive.

Chapter 8: Introduction to Linear Regression

Difficulty: Medium · Key topics: residuals, correlation, least squares regression, slope and intercept interpretation, leverage and outliers, inference for regression coefficients

This chapter introduces fitting a straight line to two numerical variables, starting from the residual — the gap between an observed value and the line’s prediction — and correlation, which measures the strength of a linear relationship on a scale from −1 to 1. It derives the least squares line, the one that minimizes the sum of squared residuals, and shows how to recover its slope and intercept directly from the means, standard deviations, and correlation of the two variables. It then covers how to interpret the slope and intercept in context, why points far from the center of the data (high leverage) can distort the fit, and how to build a confidence interval for a regression coefficient using the t-distribution.

Key Points:

• A residual is e_i = y_i − ŷ_i, the gap between an observed value and the model’s prediction for that case
• Correlation (R) always falls between −1 and 1 and measures the strength of a linear relationship
• Slope of the least squares line from summary statistics: b1 = (sy / sx) × R, and the line always passes through (x̄, ȳ)
• The slope describes the average change in y for a one-unit increase in x; the intercept is only meaningful when x = 0 is a realistic value for the data
• Points that sit far horizontally from the center of the data have high leverage and can pull the fitted line disproportionately toward themselves
• A confidence interval for a regression coefficient uses the t-distribution: b_i ± t*_df × SE_bi

Practice Tip: Fit a line to any two-column data set you have handy and plot the residuals separately — if the residual plot shows a curve or a fan shape instead of a random scatter, a straight line is the wrong model regardless of how high the correlation looks.

Common Mistake: Interpreting the y-intercept of a regression line as a meaningful real-world quantity when x = 0 is far outside the range of the actual data (e.g. a “weight at age 0” intercept), where the line’s extrapolation has no real meaning.

Important Questions:

• What does a residual represent, and how is it calculated? A residual, e_i = y_i − ŷ_i, is the difference between an observed response and the value the fitted line predicts for that same case; a large residual means the line fits that point poorly.
• What is a high-leverage point in regression, and why does it matter? A point that lies far horizontally from the center of the explanatory-variable values; because the least squares line must pass close to it, a single high-leverage point can pull the whole fitted line toward itself.

Chapter 9: Multiple and Logistic Regression

Difficulty: Hard · Key topics: multiple regression model, reference level for categorical predictors, adjusted R², model selection (backward elimination, forward selection), checking model conditions with graphs, logistic regression, logit

This closing chapter extends simple regression to many predictors at once, using a Lending Club loan-interest-rate data set as its running example. It defines the multiple regression model, explains how a categorical predictor with several levels is represented through a reference level (the one category that receives no separate coefficient), and introduces adjusted R² as a fairer way to compare models with different numbers of predictors than ordinary R². It covers two model-selection strategies — backward elimination and forward selection — and the trade-off between optimizing for prediction accuracy versus statistical interpretability. The chapter closes with logistic regression, which swaps a numerical outcome for a two-level categorical one and models the log-odds (logit) of the outcome as a linear function of the predictors.

Key Points:

• Multiple regression model: ŷ = β0 + β1x1 + β2x2 + ... + βkxk for k predictors, with coefficients estimated by software
• For a categorical predictor with k > 2 levels, software reports k − 1 coefficients; the level left out is the reference level, and the rest are relative to it
• Adjusted R² penalizes a model for adding predictors that do not meaningfully reduce residual variance, unlike ordinary R², which never decreases as predictors are added
• Backward elimination starts with every predictor and removes the least useful one at a time; forward selection starts empty and adds the most useful predictor at a time
• Use adjusted R² when the goal is pure prediction accuracy; use the p-value approach when the goal is identifying statistically significant predictors or a simpler, more interpretable model
• A logistic regression model requires independent outcomes and a predictor that is linearly related to the logit of the outcome’s probability, holding other predictors constant

Memory Tip: Statistician George Box’s line, quoted directly in this chapter — “all models are wrong, but some are useful” — is the right way to hold every multiple regression result: report a model’s known shortcomings honestly rather than treating any fitted model as the literal truth.

Common Mistake: Comparing two multiple regression models by ordinary R² alone, which always goes up (or stays the same) when a predictor is added, even a useless one — adjusted R² or a significance test is needed to tell whether a predictor actually earned its place in the model.

Important Questions:

• What is the reference level when a categorical predictor with several levels is added to a regression model? It is the one level that receives no separate coefficient; software reports k − 1 coefficients for a k-level categorical predictor, and each reported coefficient is interpreted relative to that left-out reference level.
• Why can ordinary R² be misleading when comparing multiple regression models with different numbers of predictors? R² never decreases when a new predictor is added, even an irrelevant one, so a model with more predictors can show a higher R² without actually fitting the data meaningfully better; adjusted R² corrects for the number of predictors used.

Download OpenIntro Statistics PDF (Free)

This book is free from its official source, the OpenIntro project’s own site. Click below to go to the official book page, where the free PDF, the data sets, and learning objectives for every chapter are all available — no cost, no registration required for the PDF.

How to Study This Book

Chapters 1 and 2 are not optional for anyone: Chapter 1 gives you the vocabulary (variable types, observational study vs. experiment) that every later chapter assumes without re-explaining, and Chapter 2 gives you the summary statistics and graphs that every later example is built on top of. Do not skip either one even if your course jumps straight to inference.

Chapter 3 (Probability) is genuinely optional on a tight schedule — the book says so itself. If your course is mainly about confidence intervals, hypothesis tests, and regression, you can read Chapter 3’s sections on defining probability and the Addition/Multiplication Rules for intuition and skip the random-variables and continuous-distributions sections unless your syllabus specifically covers them. Chapter 4 (Distributions) is used heavily later, particularly the normal distribution and Z-scores in Chapter 4’s first section, and the binomial distribution if your course touches proportions in depth; the geometric, negative binomial, and Poisson sections are commonly skipped on a one-semester intro syllabus.

Chapter 5 (Foundations for Inference) is the hinge of the whole book. Everything in Chapters 6 and 7 is the same confidence-interval-and-hypothesis-test framework from Chapter 5, applied to a new setting and a new distribution. If Chapter 5 feels shaky, go back to it before pushing into Chapter 6 or 7 — the formulas change, but the logic (point estimate, standard error, compare to a critical value or a p-value) does not.

Chapter 6 (categorical data) and Chapter 7 (numerical data) can largely be read in either order depending on which type of outcome variable your course covers first, but Chapter 7’s t-distribution material assumes you are comfortable with the normal-distribution reasoning from Chapter 5. Within Chapter 7, the ANOVA section (comparing 3+ group means) is the one most often reserved for a second course rather than a first one — check your syllabus before spending heavy study time there.

Chapters 8 and 9 (regression) form their own two-part unit and should be read in order: Chapter 9’s multiple regression and logistic regression both build directly on the residual, correlation, and least-squares vocabulary Chapter 8 sets up for a single predictor. If your course only covers simple linear regression, Chapter 8 alone is a complete, self-contained unit; skip Chapter 9 until a course that explicitly covers multiple or logistic regression.

This book does not cover: Bayesian statistics as a primary framework (Bayes’ Theorem appears only as one conditional-probability tool in Chapter 3), nonparametric methods, survival analysis, or time series — if your course includes any of those topics, you will need a supplementary source alongside this book.

Used In These Programs

This book is used for the Introductory Statistics course in: BS Statistics · BS Computer Science. Browse all Statistics books or all Statistics category books.

Who Should Read This

OpenIntro Statistics is written for a first course in statistics and assumes no prior statistics background and, deliberately, no calculus — it is the right starting point for BS Statistics and BSCS students taking an Introductory Statistics or Probability & Statistics course where calculus is not a prerequisite. It is not the right book for a calculus-based mathematical statistics course (one that derives distributions using integrals and expects comfort with continuous probability theory from the start), and it is not written as a data-science-first, programming-heavy text either. Students who want a tidyverse/R-centric treatment of similar material, or who are moving on to a more modern, code-first second course, should pair this book with its newer sibling, Introduction to Modern Statistics, once that page is published on this site.

Applicable Universities

This book is useful for students at Pakistani universities offering BS Statistics and BSCS degrees, including Punjab University, Virtual University, COMSATS, FAST, UET, NUST, and other HEC-recognized institutions.

FAQs

Related Books

• Discrete Mathematics: An Open Introduction – Oscar Levin (4th Edition)
• Calculus Volume 1 – Strang & Herman (OpenStax)

OpenIntro Statistics is this project’s first Statistics category book — the calculus-free Introductory Statistics course most BS Statistics and BSCS students take, covering data collection, summarizing data, probability, distributions, confidence intervals, hypothesis testing, and linear regression in one sequence. Browse more Statistics books as this category grows, or related Mathematics books for the rest of your degree.

OpenIntro Statistics, Fourth Edition, by David Diez, Mine Çetinkaya-Rundel, and Christopher D Barr. Free, licensed under the Creative Commons Attribution-ShareAlike 3.0 Unported License (CC BY-SA 3.0). Access for free at https://www.openintro.org/book/os