Getting Started with Statistics

Statistics helps you make sense of the world around you by collecting and analysing data. Whether you're looking at exam results, social media trends, or scientific research, the methods you'll learn here are the foundation for understanding what the numbers actually mean.

The key is knowing how to collect good data and then interpret it properly. You'll discover different ways to gather information, measure what matters, and spot patterns that can help predict future outcomes.

Data Collection Methods

Census vs sampling - that's your first big choice when collecting data. A census measures every single person in your population (like the UK national census), giving you perfect accuracy but costing loads of time and money.

Sampling is much more practical for most situations. You'll use a sampling frame (your list of potential participants) to select sampling units (the actual people you'll survey). Simple random sampling gives everyone an equal chance using a random number generator - it's completely unbiased but you need that complete list first.

Systematic sampling is brilliant when you want speed. Calculate k = population size ÷ sample size, pick a random starting point, then take every kth person. Stratified sampling ensures your sample represents different groups proportionally - work out (sample ÷ population) × group size, then randomly select within each group.

Quick tip: Random sampling methods eliminate bias but need complete lists, whilst non-random methods like quota and opportunity sampling are cheaper but potentially biased.

Non-Random Sampling and Data Types

Quota sampling works like stratified sampling, but researchers fill each group however they want - no random selection needed. It's dead easy since you don't need a complete sampling frame, but bias can creep in. Opportunity sampling just grabs whoever's available at the time - think surveying people outside Tesco. Super cheap and quick, but hardly representative of the whole population.

Data types matter loads for your analysis. Qualitative data describes qualities (like favourite colours or opinions), whilst quantitative data deals with actual numbers. Quantitative splits into discrete (whole numbers like number of siblings) and continuous (measurements like height that can be any value).

Understanding these differences helps you choose the right analysis methods later on. You can't calculate an average of favourite pizza toppings, but you absolutely can find the mean number of pets people own.

Remember: The sampling method you choose affects how trustworthy your results will be, so pick wisely based on your time, budget, and accuracy needs.

Measures of Location and Quartiles

Finding the mean is straightforward: add everything up and divide by how many values you have. For frequency tables, use Σfx ÷ Σf. But quartiles show you where your data splits into quarters, giving you a much better picture of the distribution.

Listed data quartiles are simple - Q₁ sits at position n/4, Q₂ (the median) at n/2, and Q₃ at 3n/4. If you get a decimal position, round up. If it's a whole number, find the midpoint with the next value.

Grouped data needs linear interpolation because your quartile position might fall somewhere within a class interval. Find which class contains your quartile position, then use the formula to work out exactly where it sits within that range.

Percentiles and deciles work similarly - the 57th percentile sits at 0.57 × n, whilst deciles split your data into 10% chunks. Never round these positions; always use linear interpolation for the most accurate results.

Key insight: Quartiles tell you more about your data's shape than the mean alone ever could - they show you whether values cluster together or spread out widely.

Spread and Variation in Data

The interquartile range (IQR) equals Q₃ - Q₁ and brilliantly ignores extreme values that might skew your results. Interpercentile ranges work the same way - P₉₀ - P₁₀ shows the spread whilst ignoring the most extreme 10% at each end.

Variance (σ²) measures how spread out your data is from the mean. Use the formula: Σx²/n - (x̄)². For frequency data, it's Σfx²/Σf - x̄². Remember, Σfx² is completely different from (Σfx)² - don't mix them up!

Coding makes calculations easier when dealing with awkward numbers. If y = ax + b, then the mean becomes ȳ = ax̄ + b. The standard deviation changes to σy = aσx. Notice that adding or subtracting values doesn't change the standard deviation - only multiplication or division affects spread.

Think "mean of squares minus square of the means" for variance. This helps you remember the correct order of operations and avoid common mistakes.

Pro tip: Standard deviation isn't affected by adding or subtracting the same value to all your data points - only scaling (multiplying) changes the spread.

Visual Data Representation

Cumulative frequency curves and box plots show your data's distribution at a glance. Box plots display the five-number summary (min, Q₁, Q₂, Q₃, max) and highlight outliers as individual points beyond the whiskers.

Histograms work for continuous data with no gaps between bars. The key is frequency density = frequency ÷ class width. This adjustment ensures areas represent frequencies correctly, even when class widths vary.

When comparing distributions, focus on three elements: location (where the centre sits), spread (how scattered the values are), and always use the context of your data. Don't just state that "Dataset A has a higher mean" - explain what this means for the real situation you're studying.

Statistical charts aren't just pretty pictures - they're powerful tools for spotting patterns, outliers, and trends that raw numbers might hide.

Visual learning: A well-drawn histogram or box plot can reveal patterns in seconds that might take ages to spot in a table of numbers.

Probability Fundamentals

Probability measures how likely events are to happen, from 0 (impossible) to 1 (certain). Union (A ∪ B) means "A OR B happens" - shade both circles in a Venn diagram. Intersection (A ∩ B) means "both A AND B happen" - shade only the overlap.

Tree diagrams map out multiple events brilliantly. Each branch shows a probability, and you multiply along branches to find combined probabilities. For independent events, P(A ∩ B) = P(A) × P(B) - one event doesn't affect the other.

Conditional probability asks "what's the chance of B happening, given that A has already occurred?" Use P(B|A) = P(A ∩ B) ÷ P(A). This is massive in real life - like finding the probability of rain given it's cloudy.

Mutually exclusive events can't happen simultaneously - if A occurs, B definitely can't. So P(A ∩ B) = 0, and P(A ∪ B) = P(A) + P(B). Think of rolling a dice - you can't get both a 3 AND a 5 on the same roll.

Real-world connection: Conditional probability explains why medical tests aren't 100% reliable and why weather forecasts give percentage chances rather than definite predictions.

Binomial Distribution and Hypothesis Testing

Binomial distribution X~B(n,p) models situations with fixed trials, fixed probability, independence, and just two outcomes $success/failure$. Use P$X=r$ = (n choose r) × p^r × $1-p$^$n-r$ to find exact probabilities.

Cumulative probabilities need careful attention to inequalities. P(X≤4) includes 4, but P(X<5) also equals P(X≤4). For P(X≥3), calculate 1-P(X≤2). Always check whether your boundary values are included.

Hypothesis testing starts with the null hypothesis (H₀) - what you assume is true. The alternative hypothesis (H₁) is what you suspect might actually be happening. Your significance level sets the threshold for deciding whether to reject H₀.

One-tailed tests look for changes in a specific direction (H₁: p>k or p<k), whilst two-tailed tests just check for any difference (H₁: p≠k). For two-tailed tests, split your significance level in half.

Critical thinking: Hypothesis testing doesn't "prove" anything - it just tells you whether your evidence is strong enough to doubt what you initially assumed was true.

Statistical Testing and Correlation

Binomial hypothesis testing uses your observed number of successes as the test statistic. Assume H₀ is true, find P(X≥ or ≤ observed value), then compare with your significance level. If P < significance level, reject H₀. If P > significance level, there's insufficient evidence to reject H₀.

Correlation vs causation - just because two variables move together doesn't mean one causes the other. Ice cream sales and drowning incidents both increase in summer, but ice cream doesn't cause drowning. Always consider lurking variables that might explain apparent relationships.

Linear regression finds the line of best fit using y = a + bx. The least squares regression line minimises the sum of squared errors, giving you the most accurate predictions possible. The slope (b) tells you how much y changes for each unit increase in x.

Understanding what your regression equation means in context is crucial. If y = 20 + 3x represents exam marks vs revision hours, then 20 is the expected mark with zero revision, and each hour of revision adds 3 marks on average.

Exam success: Always interpret your regression coefficients in the context of the problem - don't just state the numbers without explaining what they mean in real terms.

Linear Regression Applications

Linear regression only works when your data actually follows a straight-line pattern. If your scatter plot looks curved or random, linear regression will give misleading results. Always plot your data first to check this assumption.

Interpolation estimates values within your data range and is generally reliable. Extrapolation predicts values outside your data range but becomes increasingly unreliable the further you go. Avoid extrapolation in exams unless specifically asked.

Your regression line helps predict the dependent variable (y) from the independent variable (x), but only use it within the range of your original data. Making predictions about exam marks for 50 hours of revision when your data only goes up to 20 hours is asking for trouble.

The regression equation tells a story about relationships. A positive slope means variables increase together, whilst a negative slope indicates that as one increases, the other decreases. The y-intercept shows the predicted y-value when x equals zero.

Practical wisdom: Regression lines are powerful prediction tools, but they're only as good as the data you feed them - garbage in, garbage out!