Visualizing Time: Designing Graphical Representations for Statistical Data (Statistics and Computing)
Graham Wills
Language: English
Pages: 256
ISBN: B007EMNC84
Format: PDF / Kindle (mobi) / ePub
Art, or Science? Which of these is the right way to think of the field of visualization? This is not an easy question to answer, even for those who have many years experience in making graphical depictions of data with a view to help people understand it and take action. In this book, Graham Wills bridges the gap between the art and the science of visually representing data. He does not simply give rules and advice, but bases these on general principles and provide a clear path between them
This book is concerned with the graphical representation of time data and is written to cover a range of different users. A visualization expert designing tools for displaying time will find it valuable, but so also should a financier assembling a report in a spreadsheet, or a medical researcher trying to display gene sequences using a commercial statistical package.
value we produce a y value by averaging the y values of data lying within a given distance of the x value on the x dimension. Figure 2.13 shows a loess smooth applied to data. Loess [26] adapts basic regression by calculating a new regression fit for each value of x by fitting a regression line only to the data within a given distance of that value, and with decreasing weights the further away we get from that central location. For this example, the distance used is fixed at 1 year, so when we
that show different aspects of the data and lead to a full understanding. In Fig. 3.20 the goal is to discover which Beatles songs a listener likes best based on a set of available data (year, album, song length). This chart shows one aspect; answering a question on how much the ratings depend on the length of the song: Long songs are rated more highly, but the correlation is not strong. Figure 3.21 shows ratings by album, with album dates shown using color.6 A couple of albums stand out at
like a single entity, which makes the chart easier to compare to other similar charts. For the movie data, we could categorize movies into types – drama, comedy, action, etc. – and produce small multiples of the basic chart for each category. This faceted version would allow us to compare patterns between movie genres and answer questions such as whether action movies show yearly patterns similar to those of comedies or whether horror films have higher peaks or longer tails than dramas. Figure
It is a summarizing aesthetic using the sum of the counts as the summary statistic. • Color – the color of an ellipse shows the average year in which the soldiers represented by the glyph were active. It is a summarizing aesthetic using the mean of the year as the summary statistic. 2 It is generally a good idea to avoid 3-D anyway. Since we are not good at judging position in 3-D, and our perception of 3-D is highly conflated with how we perceive aesthetics such as size, brightness, and
data that make a picture look good, but rather to choose data for which a chart should be applicable. If the result is not perfect, I prefer to present imperfection and explore remedies rather than look for a different data source. vii viii Preface This book is concerned with the graphical representation of time data. Time is special – it doesn’t behave quite like other variables. It has an inherent direction and determines causality. Time can be recorded in many ways: it can be linear or