Category: Rstats

Running the Numbers – How Can Hamilton Still Take the 2016 F1 Drivers’ Championship?

Way back in 2012, I posted a simple R script for trying to work out the finishing combinations in the last two races of that year’s F1 season for Fernando Alonso and Sebastien Vettel to explore the circumstances under which Alonso could take the championship (Paths to the F1 2012 Championship Based on How They Might Finish in the US Grand Prix); I also put together a simple shiny version of the script to make it bit more app like (Interactive Scenarios With Shiny – The Race to the F1 2012 Drivers’ Championship), which I also updated for the 2014 season (F1 Championship Race, 2014 – Winning Combinations…).

And now we come to 2016, and once again, with two races to go, there are two drivers in with a chance of winning overall… But what race finishing combinations could see Hamilton make a last stand and reclaim his title? The  F1 Drivers’ Championship Scenarios, 2016 shiny app will show you…


You can find the code in a gist here:

DH Box – Digital Humanities Virtual Workbench

As well as offering digital application shelves, should libraries offer, or act as instituional sponsors of, digital workbenches?

I’ve previously blogged about things like SageMathCloud, and application based learning environment, and the IBM Data Scientist Workbench, and today came across another example: DHBox, CUNY’s digital humanities lab in the cloud (wiki), which looks like it may have been part of a Masters project?


If you select the demo option, a lab context is spawned for you, and provides access to a range of tools: staples, such as RStudio and Jupyter notebooks, a Linux terminal, and several website creation tools: Brackets, Omeka and WordPress (though the latter two didn’t work for me).


(The toolbar menu reminded me of Stringle / DockLE ;-)

There’s also a file browser, which provides a common space for organising – and uploading – your own files. Files created in one application are saved to the shared file area and available for use on other applications.


The applications are being a (demo) password authentication scheme, which makes me wonder if persistent accounts are in the project timeline?


Once inside the application, you have full control over it. If you need additional packages in RStudio, for example, then just install them:


They work, too!


On the Jupyter notebook front, you get access to Python3 and R kernels:



In passing, I notice that RStudio’s RMarkdown now demonstrates some notebook like activity, demonstrating the convergence between document formats such as Rmd (and ipymd) and notebook style UIs [video].

Code for running your own DHBox installation is available on Github (DH-Box/dhbox), though I haven’t had a chance to give it a try yet. One thing it’d be nice to see is a simple tutorial showing how to add in another tool of your own (OpenRefine, for example?) If I get a chance to play with this – and can get it running – I’ll try to see if I can figure out such an example.

It also reminded me that I need to play with my own install of tmpnb, not least because of  the claim that “tmpnb can run any Docker container”.  Which means I should be able to set up my own tmpRStudio, or tmpOpenRefine environment?

If visionary C. Titus Brown gets his way with a pitched for MyBinder hackathon, that might extend that project’s support for additional data science applications such as RStudio, as well as generalising the infrastructure on which myBinder can run. Such as Reclaimed personal hosting environments, perhaps?!;-)

That such combinations are now popping up all over the web makes me think that they’ll be a commodity service anytime soon. I’d be happy to argue this sort of thing could be used to support a “technology enhanced learning environment”, as well as extending naturally into“technology enhanced research environments”, but from what I can tell, TEL means learning analytics and not practical digital tools used to develop digital skills? (You could probably track the hell of of people using such environments if you wanted to, though I still don’t see what benefits are supposed to accrue from such activity?)

It also means I need to start looking out for a new emerging trend to follow, not least because data2text is already being commoditised at the toy/play entry level. And it won’t be VR. (Pound to a penny the Second Life hipster, hypster, shysters will be chasing that. Any VR campuses out there yet?!) I’d like to think we might see inroads being made into AR, but suspect that too will always be niche, outside certain industry and marketing applications. So… hmmm… Allotments… that’s where the action’ll be… and not in a tech sense…

Using Docker as a Personal Productivity Tool – Running Command Line Apps Bundled in Docker Containers

With its focus on enterprise use, it’s probably with good reason that the Docker folk aren’t that interested in exploring the role that Docker may have to play as a technology that supports the execution of desktop applications, or at least, applications for desktop users. (The lack of significant love for Kitematic seems to be representative of that.)

But I think that’s a shame; because for educational and scientific/research applications, docker can be quite handy as a way of packaging software that ultimately presents itself using a browser based user interface delivered over http, as I’ve demonstrated previously in the context of Jupyter notebooks, OpenRefine, RStudio, R Shiny apps, linked applications and so on.

I’ve also shown how we can use Docker containers to package applications that offer machine services via an http endpoint, such as Apache Tika.

I think this latter use case shows how we can start to imagine things like a “digital humanities application shelf” in a digital library (fragmentary thoughts on this), that allows users to take either an image of the application off the shelf (where an image is a thing that lets you fire up a pristine instance of the application), or a running instance of the application of the shelf. (Furthermore, the application can be run locally, on your own desktop computer, or in the cloud, for example, using something like a mybinder like service). The user can then use the application directly (if it has a browser based UI), or call on it from elsewhere (eg in the case of Apache Tika). Once they’re done, they can keep a copy of whatever files they were working with and destroy their running version of the application. If they need the application again, they can just pull a new copy (of the latest version of the app, or the version they used previously) and fire up a new instance of it.

Another way of using Docker came to mind over the weekend when I saw a video demonstrating the use of the contentmine scientific literature analysis toolset. The contentmine installation instructions are a bit of a fiddle for the uninitiated, so I thought I’d try to pop them into a container. That was easy enough (for a certain definition of easy – it was a faff getting node to work and npm to be found, the Java requirements took a couple of goes, and I;m sure the image is way bigger than it really needs to be…), as the Dockerfile below/in the gist shows.

But the question then was how to access the tools? The tools themselves are commandline apps, so the first thing we want to do is to be able to call into the container to run the command. A handy post by Mike English entitled Distributing Command Line Tools with Docker shows how to do this, so that’s all good then…

The next step is to consider how to retain copies of the files created by the command line apps, or pass files to the apps for processing. If we have a target host directory and mount it into the container as as a shared volume, we can keep the files on our desktop or allow the container to create files into the host directory. Then they’ll be accessible to us all the time, even if we destroy the container.

The gist that should be embedded below shows the Dockerfile and a simple batch file passes the Contentmine tool commands into the container which then executes them. The batch file idea could be further extended to produce a set of command shortcuts that essentially alias the Contentmine commands (eg a ./getpapers command rather than a ./contentmine getpapers command, or that combine the various steps associated with a particular pipeline or workflow – getpapers/norma/cmine, for example – into a single command.

UPDATE: the CenturyLinkLabs DRAY docker pipeline looks interesting in this respect for sequencing a set of docker containers and passing the output of one as the input to the next.

If there are other folk out there looking at using Docker specifically for self-managed “pull your own container” individual desktop/user applications, rather than as a devops solution for deploying services at scale, I’d love to chat…:-)

PS for several other examples of using Docker for desktop apps, including accessing GUI based apps using X WIndows / X11, see Jessie Frazelle’s post Docker Containers on the Desktop.

PPS See also More Docker Doodlings – Accessing GUI Apps Via a Browser from a Container Using Guacamole for an attempt at exposing a GUI based app, such as Audacity, running in a container via a browser. Note that I couldn’t get a shared folder or the audio to work, although the GUI bit did…

PPPS I wondered how easy it would be to run command-line containers from within Jupyter notebook itself running in inside a container, but got stuck. Related question on Stack Overflow here.

The rest of the way this post is published is something of an experiment – everything below the line is pulled in from a gist using the WordPress – embedding gists shortcode…

First Thoughts on Detecting Motorsport Safety Car Periods from Laptimes

Prompted by Markku Hänninen, I thought I’d have a quick look at estimating motorsport safety car laps from a set of laptime data. For the uninitiated, if there is a dangerous hazard on track, the race-cars are kept out while the hazard is cleared, but led around by a safety car that limits the pace. No overtaking is allowed for race position, but under certain regulations, lapped cars may unlap themselves. Cars may also pit under the safety car.

Timing sheets typically don’t identify safety car periods, so the question arises: how can we detect them?

One condition that is likely to follow is that the average pace of the laps under safety car conditions will be considerably slower than under racing conditions. A quick way of estimating the race pace is to find the fastest laptime across the whole of the race (or in an online algorithm, the fastest laptime to date).

With a lapTimes dataframe containing columns lap, rawtime (that is, raw laptime in seconds) and position (that is, the race position of the driver recording a particular laptime on a particular lap), we can easily find the fastest lap:


We can find the mean laptime per lap using ddply() to group around each lap:

ddply(lapTimes[c('lap', 'rawtime', 'position')], .(lap), summarise,
      mean_laptime=mean(rawtime) )

We can also generate a variety of other measures. For example, within the grouped ddply operation, if we divide the mean laptime per lap (mean(rawtime)) by the fastest overall laptime we get a normalised mean laptime based on the fastest lap in the race.

ddply(lapTimes[c('lap','rawtime')], .(lap), summarise,
      norm_laptime=mean(rawtime)/minl )

We might also normalise the leader’s laptime for each lap, on the basis that the leader will be the car most likely to be driving at the safety car’s pace. (The summarising is essentially redundant here because we only have one row per group.

ddply(lapTimes[lapTimes['position']==1, c('lap','rawtime')], .(lap), summarise,
      norm_leaders_laptime=mean(rawtime)/minl )

Using the normalised times, we can identify slow laps. For example, slow laps based on mean laptime. In this case, I am using a heuristic that says the laptime is a slow laptime if the normlised time is more than 1.3 times that of the fastest lap:

ddply(lapTimes[c('lap','rawtime')], .(lap), summarise,
      slow_lap_meanBasis= (mean(rawtime)/minl) > 1.3 )

If we assume that the first lap does not start under the safety car, we can then make a crude guess that a slow lap not on the first lap is a safety car lap.

However, this does not take into account things like sudden downpours or other changes to the weather or track conditions. In such a case it may be likely that the majority of the field pits, so we might want to have a detector that flags whether a certain number of cars have pitted on a lap, possibly normalised against the current size of the field.

lapTimes=merge(laps, pits, by=c('lap','driverId'), all.x=T)

#Count of stops per lap
ddply( lapTimes, .(lap), summarise, ps=sum(pitstop==TRUE) )

#Proportion of cars stopping per lap
ddply( lapTimes, .(lap), summarise, ps=sum(pitstop==TRUE)/length(pitstop) )

That said, under safety car conditions, many cars do also take the opportunity to pit. However, under sudden changes of weather condition, we might expect nearly all the cars to come in, even if it means doubling up. (So another detector for weather might be two cars in the same team, close to each other in terms of gap, pitting on the same lap, with the result that one will be queued behind the other.)

As and when I get a chance, I’ll try to add some sort of ‘safety car’ estimator to the Wrangling F1 Data With R book.

When Documents Become Databases – Tabulizer R Wrapper for Tabula PDF Table Extractor

Although not necessarily the best way of publishing data, data tables in PDF documents can often be extracted quite easily, particularly if the tables are regular and the cell contents reasonably space.

For example, official timing sheets for F1 races are published by the FIA as event and timing information in a set of PDF documents containing tabulated timing data:


In the past, I’ve written a variety of hand crafted scrapers to extract data from the timing sheets, but the regular way in which the data is presented in the documents means that they are quite amenable to scraping using a PDF table extractor such as Tabula. Tabula exists as both a server application, accessed via a web browser, or as a service using the tabula extractor Java application.

I don’t recall how I came across it, but the tabulizer R package provides a wrapper for tabula extractor (bundled within the package), that lets you access the service via it’s command line calls. (One dependency you do need to take care of is to have Java installed; adding Java into an RStudio docker container would be one way of taking care of this.)

Running the default extractor command on the above PDF pulls out the data of the inner table:

extract_tables('Best Sector Times.pdf')


Where the data is spread across multiple pages, you get a data frame per page.


Note that the headings for the distinct tables are omitted. Tabula’s “table guesser” identifies the body of the table, but not the spanning column headers.

The default settings are such that tabula will try to scrape data from every page in the document.


Individual pages, or sets of pages, can be selected using the pages parameter. For example:

  • extract_tables('Lap Analysis.pdf',pages=1
  • extract_tables('Lap Analysis.pdf',pages=c(2,3))

Specified areas for scraping can also be specified using the area parameter:

extract_tables('Lap Analysis.pdf', pages=8, guess=F, area=list(c(178, 10, 230, 500)))

The area parameter appears to take co-ordinates in the form: top, left, width, height is now fixed to take co-ordinates in the same form as those produced by tabula app debug: top, left, bottom, right.

You can find the necessary co-ordinates using the tabula app: if you select an area and preview the data, the selected co-ordinates are viewable in the browser developer tools console area.


The tabula console output gives co-ordinates in the form: top, left, bottom, right so you need to do some sums to convert these numbers to the arguments that the tabulizer area parameter wants.


Using a combination of “guess” to find the dominant table, and specified areas, we can extract the data we need from the PDF and combine it to provide a structured and clearly labeled dataframe.

On my to do list: add this data source recipe to the Wrangling F1 Data With R book…

First Thoughts on Automatically Generating Accessible Text Descriptions of ggplot Charts in R

In a course team accessibility briefing last week, Richard Walker briefly mentioned a tool for automatically generating text descriptions of Statistics Canada charts to support accessibility. On further probing, the tool, created by Leo Ferres, turned out to be called iGraph-Lite:

… an extensible system that generates natural language descriptions of statistical graphs, particularly those created for Statistics Canada online publication, “The Daily”. iGraph-Lite reads a graph from a set of graphing programs, builds intermediate representations of it in several formats (XML, CouchDB and OWL) and outputs a description using NLG algorithms over a “string template” mechanism.

The tool is a C# application compiled as a Windows application, with the code available (unchanged now for several years) on Github.

The iGraph-Lite testing page gives a wide range of example outputs:


Increasingly, online charts are labeled with tooltip data that allows a mouseover or tab action to pop-up or play out text or audio descriptions of particular elements of the chart. A variety of “good chart design” principles designed to support clearer, more informative graphics (for example, sorting categorical variable bars in a bar chart by value rather than using an otherwise arbitrary alphabetic ordering) not only improves the quality of the graphic, it also makes a tabbed through audio description of the chart more useful. For more tips on writing clear chart and table descriptions, see Almost Everything You Wanted to Know About Making Tables and Figures.

The descriptions are quite formulaic, and to a certain extent represent a literal reading of the chart, along with some elements of interpretation and feature detection/description.

Here are a couple of examples – first, for a line chart:

This is a line graph. The title of the chart is “New motor vehicle sales surge in January”. There are in total 36 categories in the horizontal axis. The vertical axis starts at 110.0 and ends at 160.0, with ticks every 5.0 points. There are 2 series in this graph. The vertical axis is Note: The last few points could be subject to revisions when more data are added. This is indicated by the dashed line.. The units of the horizontal axis are months by year, ranging from February, 2005 to January, 2007. The title of series 1 is “Seasonally adjusted” and it is a line series. The minimum value is 126.047 occuring in September, 2005. The maximum value is 153.231 occuring in January, 2007. The title of series 2 is “Trend” and it is a line series. The minimum value is 133.88 occuring in February, 2005. The maximum value is 146.989 occuring in January, 2007.

And then for a bar chart:

This is a vertical bar graph, so categories are on the horizontal axis and values on the vertical axis. The title of the chart is “Growth in operating revenue slowed in 2005 and 2006”. There are in total 6 categories in the horizontal axis. The vertical axis starts at 0.0 and ends at 15.0, with ticks every 5.0 points. There is only one series in this graph. The vertical axis is % annual change. The units of the horizontal axis are years, ranging from 2001 to 2006. The title of series 1 is “Wholesale” and it is a series of bars. The minimum value is 2.1 occuring in 2001. The maximum value is 9.1 occuring in 2004.

One of the things I’ve pondered before was the question of textualising charts in R using a Grammar of Graphics approach such as that implemented by ggplot2. As well as literal textualisation of grammatical components, a small amount of feature detection or analysis could also be used to pull out some meaningful points. (It also occurs to me that the same feature detection elements could also then be used to drive a graphical highlighting layer back in the visual plane.)

Prompted by my quick look at the iGraph documentation, I idly wondered how easy it would be to pick apart an R ggplot2 chart object and generate a textual description of various parts of it.

A quick search around suggested two promising sources of structured data associated with the chart objects directly: firstly, we can interrogate the object return from calling ggplot() and it’s associated functions directly; and ggplot_build(g), which allows us to interrogate a construction generated from that object.

Here’s an example from the quickest of plays around this idea:

g=ggplot(economics_long, aes(date, value01, colour = variable))
g = g + geom_line() + ggtitle('dummy title')

#The label values may not be the actual axis limits
txt=paste('The chart titled"', g$labels$title,'";',
          'with x-axis', g$labels$x,'labeled from',
          ggplot_build(g)$panel$ranges[[1]]$x.labels[1], 'to',
          tail(ggplot_build(g)$panel$ranges[[1]]$x.labels, n=1),
          'and y-axis', g$labels$y,' labeled from',
          ggplot_build(g)$panel$ranges[[1]]$y.labels[1], 'to',
          tail(ggplot_build(g)$panel$ranges[[1]]$y.labels,n=1), sep=' ')
if (<span class="pl-s"><span class="pl-pds">'</span>factor<span class="pl-pds">'</span></span> <span class="pl-k">%in%</span> class(<span class="pl-smi">x</span><span class="pl-k">$</span><span class="pl-smi">data</span>[[<span class="pl-smi">x</span><span class="pl-k">$</span><span class="pl-smi">labels</span><span class="pl-k">$</span><span class="pl-smi">colour</span>]])){
  txt=paste(txt, '\nColour is used to represent', g$labels$colour)

  if ( class(g$data[[g$labels$colour]]) =='factor') {
    txt=paste(txt,', a factor with levels: ',
              paste(levels(g$data[[g$labels$colour]]), collapse=', '), '.', sep='')


#The chart titled "dummy title" with x-axis date labeled from 1970 to 2010 and y-axis value01 labeled from 0.00 to 1.00
#Colour is used to represent variable, a factor with levels: pce, pop, psavert, uempmed, unemploy.&amp;amp;amp;quot;

So for a five minute hack, I’d rate that approach as plausible, perhaps?!

I guess an alternative approach might also be to add an additional textualise=True property to the ggplot() call that forces each ggplot component to return a text description of its actions, and/or features that might be visually obvious from the action of the component as well as, or instead of, the graphical elements. Hmm… maybe I should use the same syntax as ggplot2 and try to piece together something along the lines of a ggdescribe library? Rather than returning graphical bits, each function would return appropriate text elements to the final overall description?!

I’m not sure if we could call the latter approach a “transliteration” of a chart from a graphical to a textual rendering of its structure and some salient parts of a basic (summary) interpretation of it, but it feels to me that this has some merit as a strategy for thinking about data2text conversions in general. Related to this, I note that Narrative Science have teamed up with Microsoft Power BI to produce a “Narratives for Power BI” plugin. I imagine this will start trickling down to Excel at some point, though I’ve already commented on how simple text summarisers can be added into to Excel using simple hand-crafted formulas (Using Spreadsheets That Generate Textual Summaries of Data – HSCIC).

PS does anyone know if Python/matplotib chart elements also be picked apart in a similar way to ggplot objects? [Seems as if they can: example via Ben Russert.]

PPS In passing, here’s something that does look like it could be handy for simple natural language processing tasks in Python: polyglot, “a natural language pipeline that supports massive multilingual applications”; includes language detection, tokenisation, named entity extraction, part of speech tagging, transliteration, polarity (i.e. crude sentiment).

PPPS the above function is now included in Jonathan Godfrey’s BrailleR package (about).

PPPPS for examples of best practice chart description, see the UK Association for Accessible Formats (UKAAF) guidance on Accessible Images.


Accessing a Neo4j Graph Database Server from RStudio and Jupyter R Notebooks Using Docker Containers

In Getting Started With the Neo4j Graph Database – Linking Neo4j and Jupyter SciPy Docker Containers Using Docker Compose I posted a recipe demonstrating how to link a Jupyter notebook container with a neo4j container to provide a quick way to get up an running with neo4j from a Python environment.

It struck me that it should be just as easy to launch an R environment, so here’s a docker-compose.yml file that will do just that:

  image: kbastani/docker-neo4j:latest
    - "7474:7474"
    - "1337:1337"
    - /opt/data

  image: rocker/rstudio
    - "8787:8787"
    - neo4j:neo4j
    - ./rstudio:/home/rstudio

  image: jupyter/r-notebook
    - "8889:8888"
    - neo4j:neo4j
    - ./notebooks:/home/jovyan/work

If you’re using Kitematic (available via the Docker Toolbox), launch the docker command line interface (Docker CLI), cd into the directory containing the docker-compose.yml file, and run the docker-compose up -d command. This will download the necessary images and fire up the linked containers: one running neo4j, one running RStudio, and one running a Jupyter notebook with an R kernel.

You should then be able to find the URLs/links for RStudio and the notebooks in Kitematic:


Once again, Nicole White has some quickstart examples for using R with neo4j, this time using the Rneo4j R package. One thing I noticed with the Jupyter R kernel was that I needed to specify the CRAN mirror when installing the package: install.packages('RNeo4j', repos="")

To connect to the neo4j database, use the domain mapping specified in the Docker Compose file: graph = startGraph("http://neo4j:7474/db/data/")

Here’s an example in RStudio running from the container:


And the Jupyter notebook:


Notebooks and RStudio project files are shared into subdirectories of the current directory (from which the docker compose command was run) on host.