1/83
Looks like no tags are added yet.
Population
The entire group of elements you want to draw conclusions about (e.g. ALL UK listed companies)
Sample
A subset of the population that is actually studied (e.g. 200 randomly selected companies)
Census
Collecting data from EVERY element in the population
Parameter
A numerical value describing a POPULATION; fixed but usually unknown (μ, σ, π)
Statistic
A numerical value calculated from a SAMPLE to estimate a parameter; varies from sample to sample (x̄, s, p)
Element
The individual unit on which data is collected (e.g. each company in the study)
Variable
The characteristic being measured on each element (e.g. revenue)
Observation
The set of values recorded for one element (e.g. one company's revenue, employees and sector)
Experimental study
Researcher controls/assigns the treatment, ideally at random (e.g. randomly assigning staff to training vs no training)
Observational study
Researcher only observes and records; no treatment is assigned or controlled
Qualitative (categorical) data
Labels or categories, not numbers that can be meaningfully calculated (e.g. industry sector, yes/no)
Quantitative data
Numerical values that measure how much or how many
Discrete variable
Quantitative data from COUNTING; whole numbers only (e.g. number of complaints, number of employees)
Continuous variable
Quantitative data from MEASURING; any value in a range (e.g. time, weight, distance)
Nominal scale
Categories with NO natural order (e.g. area codes, industry sector, gender)
Ordinal scale
Categories WITH a natural order but gaps not measurable (e.g. credit ratings AAA to D, Poor/Fair/Good)
Interval scale
Ordered, equal gaps, but NO true zero (e.g. temperature in °C)
Ratio scale
Ordered, equal gaps, AND a true zero (e.g. journey time, revenue, weight)
Four key summary aspects of a data set
Centrality, Variability, Location (locality), Shape (symmetry)
Measures of centrality
Mean, median, mode, describe the "typical" or central value
Measures of variability
Range, variance, standard deviation, IQR, coefficient of variation
Measures of location (locality)
Percentiles, quartiles, z-scores
Mean
Sum of values ÷ number of values; sensitive to extreme values
Median
Middle value of the ordered data; not affected by extreme values
Mode
Most frequently occurring value; can have none or several; often used for categorical data
Why does variance square the deviations?
Positive and negative deviations would cancel out to zero
Standard deviation
Square root of the variance; returns spread to the original units
Interquartile range (IQR)
Q3 − Q1; spread of the middle 50%; measure of SPREAD less affected by extreme values
Q1, Q2, Q3
25th percentile, 50th percentile (median), 75th percentile
Five-number summary (boxplot)
Minimum, Q1, Median, Q3, Maximum
Empirical rule
For bell-shaped data, ~68% within ±1 SD, ~95% within ±2 SD, ~99.7% within ±3 SD
Positive (right) skew
Long tail to the right; mean > median; median closer to Q1 in a boxplot
Negative (left) skew
Long tail to the left; mean < median
Same mean, different SD, which is riskier?
The one with the HIGHER standard deviation (more variability)
Relative frequency (empirical) probability
Probability estimated from observed/sample data, number of times occurred ÷ number of trials
Subjective probability
A personal judgement used when there is no history of outcomes (e.g. a manager's 70% estimate)
Law of large numbers
As the number of trials increases, relative frequency settles towards the true probability
Experiment
An activity with an observable outcome
Trial
One repetition of an experiment
Sample space
A complete list of ALL possible elementary outcomes (e.g. sum of two dice, {2, 3, …, 12})
Elementary vs compound event
Elementary cannot be broken down (rolling a 2); compound can (rolling an even number = 2, 4, 6)
Mutually exclusive events
Cannot occur at the same time, P(A and B) = 0, so P(A or B) = P(A) + P(B)
Independent events
One does not affect the other, P(A | B) = P(A) and P(A and B) = P(A) × P(B)
Marginal probability
Probability of a single event; row or column TOTAL ÷ grand total
Joint probability
Probability of two events together; one CELL ÷ grand total
Random variable
A numerical outcome of a random process, unknown until it happens (usually X)
Discrete probability distribution
List of every possible value of X with its probability; probabilities sum to 1
Binomial conditions
Fixed number of trials n; only two outcomes per trial; constant probability of success π; independent trials (π does NOT have to be 0.5)
Binomial "success"
The outcome being COUNTED, occurring with probability π; can be a bad thing (defect, no-show)
Poisson distribution use
Count of occurrences in a fixed interval of time, space or distance, with no upper limit (e.g. calls per hour, flaws per metre)
Poisson assumptions
Constant rate across equal intervals; occurrences in different intervals are independent
Binomial vs Poisson
Binomial, fixed n, need n and π. Poisson, no fixed n, no upper limit, need λ for the interval
Continuous random variable, P(X = one exact value)
Probability density function properties
Non-negative; total area under curve = 1; probability at a single point = 0
Normal distribution properties
Bell-shaped and symmetric; mean = median = mode; spread set by σ; tails extend infinitely (no fixed min or max)
Standard normal distribution
Normal with μ = 0 and σ = 1; values called z
Setting a target only the top 10% exceed
Find the 90th percentile of the distribution
Sampling distribution
The distribution of a statistic (e.g. x̄) across all possible samples of the same size
Mean of the sampling distribution of x̄
Equal to the population mean μ
Central Limit Theorem
Sampling distribution of x̄ is approximately normal for large n, whatever the population shape
CLT rules of thumb
Normal population, any n. Symmetric population, n ≥ 15. Unknown shape, n ≥ 30
Conditions to use normal for a proportion
nπ ≥ 10 and n(1 − π) ≥ 10
Point estimate
A single-value best guess of a parameter (x̄, p, s)
Correct meaning of a 95% confidence interval
Intervals built this way capture the parameter in 95% of samples. Say "95% confident", NOT "95% chance μ is in it"
Effect of higher confidence level
Bigger multiplier, so WIDER interval
Effect of larger sample size on a CI
Smaller SE, so NARROWER interval
When to use t instead of z
σ unknown and population normal (or n ≥ 30)
Why t intervals are wider than z
Using s instead of σ adds extra uncertainty, so t > z
Hypothesis
A claim about a population parameter
Null hypothesis (H0)
The status quo assumed true; always contains the equals part (=, ≤ or ≥)
Alternative hypothesis (H1)
The competing claim; uses ≠, < or >
Two-sided test
H1 uses ≠; 5% level critical values ±1.96
One-sided test
H1 uses < or >; 5% level critical value 1.645 (or −1.645 for lower tail)
p-value
Probability of a sample result at least as extreme as the one observed, assuming H0 is true
Decision rule using p-value
Reject H0 if p-value < significance level (e.g. 5%). "If the p is low, the null must go"
Significance level (α)
Risk of a Type I error you are willing to accept
Type I error
Rejecting H0 when H0 is actually true (probability α)
Type II error
Failing to reject H0 when H0 is actually false (probability β)
Correct wording when the result is not significant
"Do not reject H0". Never "accept H0"
Classical probability
Used when the population is countable, its composition is known and selection is unbiased (dice, cards, coins)
Conditional probability
The probability an event occurs GIVEN that another event has already occurred
Bayes' rule (concept)
Revising a prior probability with new information to get a posterior probability
Range (concept)
Largest minus smallest value; very sensitive to extreme values
Discrete uniform distribution
Every possible value is equally likely (e.g. rolling a fair die)