I have described the swine flu data in the previous post, now I’m going to cover the presentation of the results in my paper. The results are a set of posterior distributions. The application details do not matter here. Here is the old picture from my master’s thesis: thesis Posterior distributions were visualized with histograms. The bins in the histogram were not predefined: they were shifted so that the mode of the original distribution would be exactly in the middle of the bin. I used different colors to emphasize that the figure show three parameter families. Now I would sort the rows according to some meaningful attribute and add a grid. Here is a picture from my previous paper: Estimating the burden of A(H1N1)pdm09 influenza in Finland during two seasons (Epidemiology and Infection) 2 The shown distributions are actually histograms with 1-pixel wide bins. To achieve this level of smoothness I had to run my computation methods for a really long time. The posterior means are emphasized and labelled. The prior distributions are presented at the top of each panel. The difference between the prior and the posterior shows how much information is gained thought the modelling. My favorite design is in the following figure: the height of each histogram is proportional to the square root of the maximum of the distribution. This is made to emphasize more certain estimates (see principles of posterior visualization) I have tried to improve the collection of histograms even further in my current paper. Here is one of the first publishable version of the figure with the results: 4 I removed the emphasis and labels from the mean values. This produced too much visual noise and leaved an overconfident impression about the certainty of the estimates. The picture does not need to present all the numbers, a tables or a text would be a better media for that. After some consideration, I decided to move from the histograms to the concentration-histograms: 5 In the concentration histograms the probability density is represented not with the height of the bin, but with the concentration of color. I believe these figures deliver the idea of uncertainty much better. In a manner of the previous picture, the maximal concentration of color is proportional to the square root of the maximum of the distribution. The order of panels is reorganized. I decided to remove the panel captions and put the colors back. F_Final_pqsg2 Stacked histograms in one of the subfigures create optical illusion: Fig_7_z - копияTo remove the illusion, I decided to add grid lines. Also, axes were almost invisible on print, I had to make them wider. Here are the final results: Fig_7


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s