1. Tables, graphs, and diagrams

2. Tables, graphs and diagrams….
organise
summarise
display
...data from outbreaks, surveillance and studies
Firstly, you need to organise your data. They may be totally chaos, for instance a bunch of paper bits with patients name and characteristics. So, you sort them by sex, age, nationality and so on.
Then you may want to organise your data. In stead of 50 lines of names with age and sex and nationality, you can show in a table that there are 20 women and 30 men and so on.
Finally, you want to display your data. By way of the design of your graphics, you want your data to be easily understood and interpretable. You want your message to speak through the design.
And the data can be any data. For us, it’s mostly data from the ordinary surveillance or form some special study.

3. Uses of tables, graphs and diagrams
Effective presentation
to the professionals
to the public
Easy interpretation
Visual examination
Analysis
Data graphics are used very often, maybe too often. With modern computer programmes, anyone - even journalists -can make professional graphics.
However, you should be conscious about why you use graphics.
It can be to make your written or oral presentation more effective; or it can be to help you interpret your data. You describe your data by the grahics and examine your graphics to look for trends, patterns, aberrations, differences and so on; that is to generate hypotheses.

4. Paper vs. screen Paper Screen Time unlimited Repetition Details
White, grey and black
Screen
Time < 1 min
No repetition
Less details
Colours possible

5. Tables, graphs, and diagrams
Self-explanatory
Simple!
Title
what, who, where, when
Define abbreviations and symbols
Note data exclusions
Reference the source
OK, then let’s make som graphics. First the basic rules that go for all graphics:
Self-explanatory means that your audience should be able to understand the graphics without reading the rest of the article or listening to you. Everything she needs should be there.
The simplest graphics are the most effective. Resist the temptation to stuff everything in to one graph. If you have two point to make, then make two graphics. And don’t let the computer decide what the graph should look like! More on this tomorrow, by Denis.
The title should include the important information: what disease is this? Which population, what time and where?
And if your data comes from someone else, say so, so that the interested reader can check the sources if he needs to.

6. Type of variables Quantitative Discrete Continous
counts, dates, cases,...
Continous
height, Hb, ….

7. Type of variables Qualitative Dichotomous Nominal Ordinal
sex, ill/not ill, ….
Nominal
religion, nationality, eye colour,...
Ordinal
social class, cancer stage, ….
Before we go on to the actual graphics, let’s agree on some types of variables, or measurement scales.
Dichotomous variables divide the world in two mutually exclusive categories. Either you have it or not, you can’t have both. For instance, either you’re ill or you are not. You may be French, or non-French. Or sex, you may have it, or not. Dichotomous variables are very common in epidemiology.
Nominal variables are divided in more than two mutually exclusive categories that have no order. You can only be one of them, but none of them are better or before the others. There is no order.
Ordinal variables are also mutually exclusive, but they have an order. One category is higher than the next, but the distance between them is not natural or numerical.
Discrete variables also have an ordering, but here the difference is the same between any two variables on any plage along the scale. For instance, with a count of cases, case 17 is one more than case 16, just as case 4 is one more than case 3.
Continuos variables can take on any value.

8. Tables

9. Types of commonly used tables
One-variable tables
Frequency distribution
Multivariable tables
Contingency tables
2x2 tables
Tables are the most used graphics. Use them for summarising data or for showing relationships among variables.
The entries are usually counts of cases, or rates or proportions.
The simplest tables is the frequency distribution of one variable, for instance number of cases in each age-group or each country.
Multivariable tables are usually only two variables. If you have more, make more tables. They are contingency tables when all entries are classified by each of the variables in the table. In epidemiology we often use the two-by-two table were the persons are classfied along two features: exposed or unexposed and ill or well. 12

10. Tab 1. Distribution of cases of salmonellosis (n=65) by age group
Tab 1. Distribution of cases of salmonellosis (n=65) by age group. Hospital A, August 2010
This is the typical frequency distribution.
Note the title: disease, person, place, time.
The labels are clear.
We have horizontal lines, but not vertical.
There is an unknown category.
We can all understand this table.
Any negative things?
Well, many people feel uncomfortable about using percents for small numbers. For instance, many journals would just cut the last column here. 8

11. Table 2. Gonorrhoea by age-group and sex, 2010
Here, we don’t give percents.
This is a bivariable contingency table.
Again, each row and column are clearly labelled.
The source of the data are given at the bottom.
Note the table number. This is like you would find in a journal, but they are usually not used in oral presentations. When writing for a journal, check a recent issue to find out how you should draw the tables.
Any negative?
Well, some of you may wonder whether the cells entries are counts or rates. It’s difficult without knowing something about the subject. Maybe the title should have said Number of cases. 7

12. Tab. IV Gastrointestinal illness and fish consumption among customers at ”X ”, 2014
Cases
Controls
Total
OR (CI95%)
Ate fish 34 20 54
13 ( )
Did not eat fish 8 62 70
Ref 42 82
124
Here’s a two by two table from a case control study.
There were 42 sick guests and 34 of them had eaten fish. That gives odds of 34:8.
Among 82 well customers you interwieved, only 20 had eaten fish. The odds was then 20:62.
So the odds for cases were 13 times as high.

13. In tables... Labels for rows and columns
Totals for rows and columns, usually
Units of measurements
Max five variables
Horizontal lines OK, vertical not
And here’s the summing up of tables.

14. Graphs & Diagrams

15. Types of commonly used graphs
Line graph
Arithmetic scale line graph
Semi-logarithmic scale line graph
Histogram
Epidemic curve
Graphs are used to assist the reader to find patterns, trends, aberrations, similarities and differences in data.
Simplicity is the most important feature.
Line graphs are used mostly for time data. They are most useful for long series of data and for comparing several data sets, for instance the development of two diseases in the same time period.
Arithmetic graphs have equal quantities along the y-axis.
Scatter plots and survival curves are used for special purposes. I want get back to those. 12

16. The arithmetic-scale line graph 1
This is also a clearly titled and labelled graph.
In my view, grid lines should almost never be used because they are not necessary, and they distort the image.
In this graph you could insert some text and an arrow pointing to this peak. What happened here?
What’s the problem here?
The rate of deaths varies between 0 and 1, while the incidence rate varies between 0 and 10. This makes it impossible to see the trend in the mortality rate. And after 1965 it’s impossible to see anything.

17. The arithmetic-scale line graph 2
For time series
Show actual changes in magnitude
X-axis = time
Y-axis = incidence (or number) of cases
Start at 0
clearly marked

18. The semilogarithmic-scale line graph 1
Here, we have changed the y-axis to a logarithmic scale. This means that the axis in divided into cycles with each cycle being ten times greater than the previous one.
So here we can draw a wide range of values, from less than 0.1 to almost 10 and still see the lines clearly.
In a semilog graph we concentrate on the rate of change. How fast does it change? A straight line is a constant rate of change. The slope indicates the the rate of change. Here, we see that the mortality changed as rapidly as the incidence.
Semilog graphs often have grid lines. It’s not necessary, and I don’t like it.

20. In graphs... Labels for axes, scales and legends
Legends or keys if >1 variable
Scale divison, appropriate scale
Units of measurements in title
No grid, no numbers
No 3D
So, let’s sum up the attributes of graphs:
Don’t use numbers in the actual graph. If you have to use numbers, make a table in stead.
And another thing, no 3D.
Keep it simple.

21. Types of commonly used charts
Charts based on length
Bar charts (horizontal, vertical, grouped, stacked)
Charts based on proportion
Component bar chart
Pie chart
Now, on to charts.
They are used for visually comparing magnitudes of an event. So, in stead of comparing the number 33 to the number 12, you compare a bar which is for instance 3.3 cm to a bar that is 1.2 cm.
In proportion charts, you divide your bar or your pie in to proportions that are the same as the proportions of the variable in each category. 21

22. Grouped bar chart

23. Stacked bar chart

24. Component bar chart

25. Bar charts Order Vertical or horizontal Same width of bars
Natural
Decreasing or increasing
Vertical or horizontal
Same width of bars
Length = frequency
Space between bars and groups, but not within groups
Tables are often better

26. Pie chart

27. The area dot (or dot density) map
Figure 2. Cases of meningococcal disease in Dublin 2006 by area of residence.
1 dot = 1 case

28. The area map
Figure 3. Incidence per 100,000 of meningococcal disease in Dublin 2006 by area of residence.

29. Every bit of ink should have a reason
Think data-ink
Data ink is the ink that is used for showing the actual data. Data ink is the part of ink that cannot be erased. It is essential. New ink should bring new information.
In a good graphic, the ratio of data ink to total ink should be as large as possible.
Every bit of ink should have a reason

30. Designing graphics Show the data Use ink for the data
Remove unnecessary ink
Remove gimmicks
No 3D
Careful with colours

35. Summary Use graphics to explore and present data
Think of difference between paper and screen
Think of your message and choose the graph type accordingly
Save your ink!