STA 490H1S: Tabular and Graphical Display of Data Alison L. Gibbs Department of Statistics University of Toronto Winter 2011 Gibbs STA 490H1S Notes from Altman and Bland I Continuous data: I I I Summary statistics such as means should not be given to more than one extra decimal place over the raw data. Greater precision may be warranted for measures of variability such as the standard deviation or standard error as they are often used in further calculations. Categorical data: I I I I Summarize as frequencies or percentages. Percentages alone can be confusing if the denominator is unclear. Percentages should usually be given as integers; one decimal place may sometimes be reasonable but not in small samples. There is rarely a need to show categorical data graphically. Gibbs STA 490H1S Notes from Altman and Bland 2 I Test statistics, correlation coefficients, and p-values: I I I No more than two decimal places for test statistics and correlation coefficients. p-values to one or two significant figures. Tables versus figures: I A figure that displays only two means with their standard errors or confidence intervals is a waste of space. Gibbs STA 490H1S Ehrenberg’s rules for tables (with contributions from Wainer) 1. Round to two significant or effective digits. Humans cannot understand more than two digits very easily. And who cares about more digits? “This year’s school budget is $27,329,681.” 2. Frame the display with suitable summary statistics. 3. Figures are easier to compare in columns than rows. 4. Order by size or some other way that makes sense. 5. Use spacing to aid perception. 6. Graphs versus tables: Graphs are of little use in communicating the quantitative aspects of the data but they can highlight qualitative results. Gibbs STA 490H1S An example from Wainer Version 1: Battery Life in Hours Battery Brands Constant Charge PowerBat Servo-Cell Never Die Electro-Blaster Cassette Player 5 7 4 8 10 Gibbs Radio 19 24 21 28 26 Flashlight 10 13 12 16 15 STA 490H1S Portable Computer 3 5 2 6 4 An example from Wainer Version 2: Battery Brands Never Die Electro-Blaster PowerBat Servo-Cell Constant Charge Radio 28 26 24 21 19 Battery Life in Hours Cassette Portable Flashlight Player Computer 16 8 6 15 10 4 13 7 5 12 4 2 10 5 3 Gibbs STA 490H1S An example from Wainer Version 3: Battery Brands Never Die Electro-Blaster PowerBat Servo-Cell Constant Charge Usage averages Radio 28 26 24 21 19 24 Battery Life in Hours Cassette Portable Flashlight Player Computer 16 8 6 15 10 4 13 7 5 12 4 2 10 5 3 13 7 4 Gibbs STA 490H1S Battery Averages 15 14 12 10 9 12 An example from Wainer Version 4: Battery Brands Never Die Electro-Blaster PowerBat Servo-Cell Constant Charge Usage averages Radio 28 26 24 21 19 24 Battery Life in Hours Cassette Portable Flashlight Player Computer 16 8 6 15 10 4 13 7 5 12 10 13 Gibbs 4 5 7 STA 490H1S 2 3 4 Battery Averages 15 14 12 10 9 12 An example from Wainer An alternative? 25 battery 15 10 5 Battery Life (hours) 20 Never Die PowerBat Electro-Blaster Constant Charge Servo-Cell Radio Flashlight Cassette Computer Appliances Gibbs STA 490H1S Burn’s ACCENT Principles for effective graphical display The essence of a graph is the clear communication of quantitative information. I Apprehension Ability to correctly perceive relations among variables. I Clarity Ability to visually distinguish all the elements of a graph. I Consistency Ability to interpret a graph based on similarity to previous graphs. I Efficiency Ability to portray a possibly complex relation in as simple a way as possible. I Necessity The need for the graph, and the graphical elements. I Truthfulness Ability to determine the true value represented by any graphical element by its magnitude relative. Gibbs STA 490H1S Pie Charts? Edward Tufte: “Pie charts are bad and that the only thing worse than one pie chart is lots of them.” Gibbs STA 490H1S Pie Charts? Edward Tufte: “Pie charts are bad and that the only thing worse than one pie chart is lots of them.” Studies have shown that piecharts are hard to read if you actually have to answer questions about the numbers they represent. This is mostly because differences in angles are not easy to judge for the human eye. Gibbs STA 490H1S References I D.G. Altman and J.M. Bland (1996) Presentation of numerical data. BMJ 312, 572. I D. A. Burn (1993), ”Designing Effective Statistical Graphs”. In C. R. Rao, ed., Handbook of Statistics, vol. 9, Chapter 22. I A.S.C. Ehrenberg (1977) Rudiments of Numeracy. Journal of the Royal Statistical Society A 140, 277-297. I A.S.C. Ehrenberg (1978) Graphs or Tables? The Statistician 27, 87-96. I H. Wainer (1997) Improving Tabular Displays, With NAEP Tables as Examples and Inspirations. Journal of Educational and Behavioral Statistics, 22, 1-30. Gibbs STA 490H1S
© Copyright 2026 Paperzz