O tipo de gráfico certo (EN)
Data de publicação: Dec 09, 2020 4:59:54 PM
You need to choose the right sort of chart for your data: the final arbiter is not how pretty it looks, but how effectively it tells the story of your data. In part this will depend on whether your data is categorised or purely numeric.
In particular, you need to be clear about the types of data you are working with. Some graphs plot numerical values for categorised data, whereas some plot two sets of related numerical data.
Categorised data is often plotted using bar, column or pie charts, but related numerical data usually requires a scatter graph.
I think we'd better look at some examples:
Categorised data
The number of fruit items eaten per month:
Fruit are categorically different items. So a bar chart or column chart like this makes sense. We've grouped the columns together by month which makes comparison a bit easier. Maybe we could use stacked bars instead, to give a better indication of the proportions of each fruit in our monthly diet. And maybe our choice of colour could better match the fruit in question (perhaps red for apple, orange for, well, orange, and green for pear?).
Numeric data
The mass of a dog and the number of biscuits eaten:
This traditional scatter plot pits one numeric value against another to see if there's a correlation. There seems to be. Maybe we'd want to join the dots to hammer this home, but a line graph tends to imply a continuum (usually a temporal one) and that's not really the case here — it's just a lot of dogs. Maybe a trend line would help though?
Implied numeric values
Some data may appear to be categorised, but is better understood as a special case of numeric data. The most common case of this is when a value is plotted over time, be that days of the week, months, or years. In this case, the time/date may need to be recorded in a format that provides a numeric value (as spreadsheets generally does with dates) so as to achieve a linear scale.
In this example, if the days are used as categories, a non-linear scale for the week is generated, but treating the days as dates includes the missing days and provides a linear scale for the horizontal axis:
Which of these is the more appropriate plot for your data? Should Wednesday have had a zero value rather than an implied value of about 50? Would point markers help in that regard? Should this be a line graph at all?
Think about what it is your chart is meant to be conveying. Don't rely on what chart comes 'out of the box'; instead, modify the chart to make it communicative and (perhaps most importantly) honest. A misleading chart is a bad chart.
Customising charts in Google Sheets
Double-click selecting a chart re-opens the Chart editor panel. When a chart is selected, the ⋮ menu allows other common actions, including moving the chart to its own sheet, downloading the chart, or applying alt text:
Retirado de York Univ.