Multiple timelines

This series of examples come from “A few simple steps to improve the description of group results in neuroscience” by Guillaume A. Rousselet, John J. Foxe and J. Paul Bolam. Rousselet, Foxe, and Bolam (2016) In this example, from xxx, patients were subjected to…

Plotting only the for each stimulus is informative, but lacking, in that we would like to know about either the spread of the data or the reliability of the estimate. Here, we opt for the 95 percent CI. In addition, the real interesting part of this experiment is at what point to the responses from the two stimuli differen enough so that we can reject the null hypothesis in a t-text. the rug below the two lines uses alpha blending to indicate where we have very low p-values.

Figure 1: A line plot of means.

Figure 2: A line plot with rug.

As an extension of this we may want to plot all the individual trend, or in this case the difference between stimuli responses per patient. The mean of all the trend lines is what’s important, but it’s nice to see all the raw data in the background. Here a line plot showing the differences with the average over-layed.

Now, if we wanted to see the difference we can just plot that, instead of making the viewer do the extra work.

Figure 3: A line plot with mean.

Having many trent lines in this regard lends itself well to being a heatmap. Here I have used the viridis colour scale to emphasis the peak values.

If we have this many time series, which are inherently quite noise, it doesn’t really make sense to show them in one big plot. An alternative would be a heat map. Although mostly frowned upon, here it works really well, in particular with the viridis color palette, since we see a really nice clear pattern.

Figure 4: A heatmap.