Data summaries, histograms and interpretation
Introduzido| English | Significado |
|---|---|
| summary statistics | single numbers that describe a data set, such as the mean, median and standard deviation |
| central tendency/ˈsentrəl ˈtendənsi/ | Where the centre of a distribution lies, measured by mean, median or mode. |
| dispersion/dɪˈspɜːʃn/ | how spread out the values of a data set are |
Four data sets, one set of statistics, 1973
- In 1973 the statistician Francis Anscombe published four small data sets. All four had the same mean of $x$, the same mean of $y$, nearly the same variances, the same correlation (0.82) and the same regression line. Plotted, they looked completely different: a scatter, a curve, a line with one outlier, and a vertical cluster with one far point.
- Summary statistics 汇总统计量 compress data — and can hide its shape. Always pair a measure of central tendency 集中趋势 and of dispersion 离散程度 with a picture.
Anscombe's four data sets had the same correlation but very different plots.
Dificuldade 1/5 · Moedas 100 + pequeno bônus na primeira conclusão hoje
Summaries can hide shape.
Centre
- Mean $\bar x = \dfrac{\sum x}{n}$; median: the middle value; mode: the most frequent.
- The mean is pulled by extreme values; the median is not.
Find the median of 3, 8, 1, 9, 4, 6.
Dificuldade 2/5 · Moedas 140 + pequeno bônus na primeira conclusão hoje
Sorted 1, 3, 4, 6, 8, 9; (4 + 6)/2.
Spread
- Range; interquartile range $\text{IQR} = Q_3 - Q_1$; standard deviation $\sigma = \sqrt{\dfrac{\sum (x - \bar x)^2}{n}}$.
- An outlier is a value below $Q_1 - 1.5\,\text{IQR}$ or above $Q_3 + 1.5\,\text{IQR}$.
Data summaries, histograms and interpretation
Compare an appropriate average and spread in context
Choose a centre and a spread measure for the worked observations.
Q₁ = 12 and Q₃ = 20. Above what value is a data point an outlier?
Dificuldade 2/5 · Moedas 140 + pequeno bônus na primeira conclusão hoje
20 + 1.5 × 8.
Pictures of data
- Histograms show the shape of grouped continuous data; box plots compare medians, quartiles and outliers; cumulative frequency curves give median and quartiles.
- Skew: a long right tail pulls the mean above the median.
In data with a long right tail, the mean is usually
Dificuldade 1/5 · Moedas 100 + pequeno bônus na primeira conclusão hoje
The tail pulls it up.
Worked example: summarising 2, 4, 4, 6, 9
- Centre: mean $= \dfrac{25}{5} = 5$; median $= 4$; mode $= 4$.
- Spread: deviations $-3, -1, -1, 1, 4$; squares $9, 1, 1, 1, 16$, sum $28$; $\sigma = \sqrt{28/5} \approx 2.37$.
- Effect of a change: adding 10 to every value adds 10 to the mean and median but leaves $\sigma$ unchanged; doubling every value doubles both mean and $\sigma$.
Find the population standard deviation of 2, 4, 4, 6, 9 to 2 d.p.
Dificuldade 2/5 · Moedas 140 + pequeno bônus na primeira conclusão hoje
√(28/5).
Every value has 10 added. What happens to the standard deviation?
Dificuldade 1/5 · Moedas 100 + pequeno bônus na primeira conclusão hoje
Spread does not move.
Choosing measures
- Skewed data or outliers: report median and IQR. Symmetric data: mean and standard deviation.
- Compare groups on both centre and spread, in context.
Which measures suit house prices with a few very expensive homes?
Dificuldade 1/5 · Moedas 100 + pequeno bônus na primeira conclusão hoje
Resistant to outliers.
Equal statistics do not mean equal data. Anscombe's four sets share their means and correlation yet differ completely in shape. Plot the data before trusting a summary.
Summarise data with a centre (mean, median) and a spread (IQR, standard deviation), chosen to suit its shape. Identify outliers with $1.5\,\text{IQR}$, use histograms and box plots to see shape, and know how adding or scaling changes each statistic.
Class A test marks: median 62, IQR 10. Class B: median 58, IQR 24. Compare the classes in context.
Centre, spread, context.