Legible Plots with Matplotlib and LibreOffice

Have you ever struggled presenting scientific plots in LibreOffice? A lot of the time the fonts come out too small or it can happen that thin lines disappear in scaled figures. Often plots are not originally created to be shown on slides. A pragmatic solution to create more legible plots with Matplotlib for LibreOffice slides is to take the dimensions that a plot should have on the final slide into account when creating the plot.

LibreOffice offers 12 standard layouts, as can be seen below:

Not all of these layouts are suitable to present plots, but for those who are, the dimensions of the layout boxes should be taken into account to create legible plots. E.g. in Matplotlib this can be accomplished with the following command:

matplotlib.rc('figure', figsize= [width,height]) 

Moreover, it is advantageous to modify some other style options to make a plot more legible. Some style options which I found helpful in the past are the following:

matplotlib.rc('lines', linewidth=4.0) #thick lines
matplotlib.rc('font',  size=18) #big font
matplotlib.rcParams.update({'font.sans-serif': 'Liberation Sans'}) #adjust font w/ slide font

In my opinion .png is one of the few trustworthy output formats for plots. Generally producing plots in vector formats (e.g. .svg) would be nicer, but at least I have experienced these being rendered incorrectly (LibreOffice 4.3.6.2 was used at the time of this writing).

An example slide with these style options applied is shown below:

The full script that was used to create the sample plots can be downloaded here. It can also be used to create sample plots for the other LibreOffice layouts.

Medisana ViFit connect Review

In the post I want to share some experiences with the Medisana ViFit connect, which is an activity tracker. It counts steps and bins them in intervals of 15min. Memory of the device is sufficient for 15 days of recording and battery lifetime is around 6 days. It has an OLED display that saves you from having to wear an additional wrist watch.

Medisana outlines that for optimal usage the activity tracker should be wrist worn. I disagree with this as my intuition is that the activity tracker overestimates the number of steps when wrist worn, especially when you spend a big portion of the day sitting.

Bluetooth 4.0 is used to upload recorded data to a smartphone, tablet, etc.. The synchronization is slow and not always successful. The app does not build an offline database, so internet connection is required while synchronizing. Moreover, synchronization always has to be explicitly initiated and is not executed as a background service.
The activity tracker has a Micro USB port for recharging.  Interestingly, hooking it up to a Linux PC loaded a driver for this USB to UART bridge. Maybe this leaves some potential for hacking? It would be great to be able to download the activity data without having to upload them to the cloud first. Another idea would be to reduce the binning interval of the activity tracker to eg. 1min to be able to draw conclusions about the types of activity (eg. running, biking, hiking, sitting, etc.) from the recorded data.

In my opinion the accompanying website cloud.vitadock.com fails to provide much insightful information (Maybe there is a legal requirement that prevents them from interpreting the activity data for the user?). It is possible to export the activity data as .csv files and I find it helpful to create two additional plots as follows:
Firstly, a histogram showing the averaged activity for all days that usually falls within the individual times of the day. An exemplary plot is shown in Fig. 1 based on mock data. It is also interesting to compare working days and weekend days in this plot.

Fig. 1.: Relative activity over time of the day

Secondly, a scatter plot showing the correlation between sleep time and the active time during the next day as shown in Fig. 2. Again this plot is based on mock data.

Fig. 2.: Active time versus sleep time