Creating a spectrogram is a basic step in every analysis of audio signals. Spectrograms visualise how frequencies change over a time period. Luckily, there is a selection of R packages that can help with this task. I will present a selection of packages that I like to use. This post is not an introduction to spectrograms. If you want to learn more about them then try other resources (e.g. lecture notes from UCL).
The examples shown below came mostly from the official documentation and were kept as simple as possible. The majority of functions allow further customisation of the plots.
seewave and ggplot2
Creating a spectrogram from the scratch is not so difficult, as shown by Hansen Johnson in this blog post. Another solution was provided by Aaron Albin.
Praat is a workhorse of audio analysis. It is a standalone software, but there is also an R controller called PraatR, that allows calling Praat functions from R. It is not the easiest tool to use so I will just mention it here for reference.
I am pretty sure that there are more packages that allow creating spectrograms but I had to stop somewhere. Feel free to leave comments about other examples.
Niedawno odnalazłem ciekawy pakiet geofacet, który umożliwia rozmieszczenie wykresów zgodnie z ich pozycją na mapie. Główna funkcja facet_geo() zastępuje facet_wrap() z ggplot2. Polska mapa jeszcze nie jest dostępna w standardowym pakiecie geofacet, ale mam nadzieję, że już wkrótce tam się znajdzie, bo dodałem ją na GitHubie.
Stworzyłem siatkę z koordynatami poszczególnych województw. Wykresy z pakietem geofacet mogą wyglądać tak:
Rozmieszczenie województw nie jest idealne, ale pakiet geofacet umożliwia użycie własnych ustawień.
Dane pochodzą z Banku Danych Lokalnych (XLS – tablica przestawna)
Kod do stworzenia wykresów:
The easiest way to plot ETG-4000 data in R is by using
plot_ETG4000() from fnirsr package. However, if you want to explore your data in more detail, then an interactive plot is more appropriate.
I used dygraphs package to create the chart below. In case of using many channels, the colours in the legend can get a bit mixed up like in my example. I haven’t figured out yet how to add a custom colour palette that could deal with multiple channels.
One way or another, this code snippet should be enough to start generating interactive charts. I haven’t added the interactive chart to the main plotting function (i.e.
plot_ETG4000) but I might do it in future releases.
The code used to generate the chart is here:
PS: The dygraph generated correctly in the interactive window, when using R notebooks, and when knitting. When I Saved as Web Page from RStudio, I got a header error
that I had to clean by removing a tag (
<!DOCTYPE html>) from the generated html file.
Recently I started to learn how to use Hitachi ETG-4000 functional near-infrared spectroscopy (fNIRS) for my research. Very quickly I found out that, as usual in neuroscience, the main data analysis packages are written in MATLAB.
I couldn’t find any script to analyse fNIRS data in R so I decided to write it myself. Apparently there are some Python options, like MNE or NinPy so I will look into them in future.
ETG-4000 records data in a straightforward(ish) .csv files but the most popular MATLAB package for fNIRS data analysis (HOMER2) expects .nirs files.
There is a ready-made MATLAB script that transforms Hitachi data into the nirs format but it’s only available in MATLAB. I will skip the transformation step for now, and will work only with a .nirs file.
The file I used (Simple_Probe.nirs) comes from the HOMER2 package. It is freely available in the package, but I uploaded it here to make the analysis easier to reproduce.
My code is here:
This will produce separate time series plots for each channel with overlapping triggers, e.g.:
The entire analysis workflow:
– RMarkdown file
– html report
I hope this helps.
More to follow.
As a fan of both open data and ROH I was excited to find the article promising that the ROH’s data might be opened. However, this article was published in 2014 and the only open data about ROH that I could find was the Royal Opera House Collections Database. Its format is far from being user-friendly, i.e. there are no cleaned csv (or even xlsx) files, but luckily the structure of the entire website is fairly predictable. This meant I managed to write a basic scraper to extract the high level performance data. Unfortunately the database only contained data for the years 1946-2012.
The result is this interactive dashboard created in Tableau:
The number of performances in 1993 (159) started to approach the peak previously reached in 1951 (168 performances).
1998 and 1999 were years when ROH was being reconstructed.
The matinees started to become more popular in the noughties.
1968 was the last year of performances as the Covent Garden Opera Company.
Tosca was the most popular opera in the studied period with 187 performances.
La bohème was the most popular matinee.
The Kirov Opera was the third most popular company performing at the ROH with 33 appearances. After the name change to Mariinsky Theatre, there were additional four performances by this company.
Here’s the scraper code:
I found a new dataset about UK broadband speeds and I started analysing it in R. However, after cleaning the data, I thought that creating a dashboard with Shiny would take me too much time so I moved to Tableau. I wanted to keep my analyses in one place so I embedded the dashboard into the output html document (see below).
Initially I thought that RMarkdown can’t generate embedded Tableau visualizations because the iframe in my report seemed blank after knitting the report. I had to open the generated in the browser to see the iframe filled with Tableau dashboard.
RMarkdown file is available here.
While working with UK geographical data I often have to extract geolocation information about the UK postcodes. A convenient way to do it in R is to use geocode function from the ggmap package. This function provides latitude and longitude information using Google Maps API. This is very useful for mapping data points but doesn’t provide information about UK-specific administrative division.
I got fed up of merging my list of postcodes with a long list of corresponding wards etc., so I looked for smarter ways of getting this info.
That’s how I came across postcodes.io which is free, open source, and based solely on open data. This service is an API to geolocate UK postcode and provide additional administrative information. Full documentation explains in details many available options. Among geographic information you can pull using postcodes.io are:
Westminster Parliamentary Constituency
European Electoral Region (EER)
Primary Care Trust (PCT)
Parish (England)/ community (Wales)
I conduct most of my analyses in R so I developed wrapper functions around the API. Developmental version of the PostcodesioR package can be found on GitHub and documentation is here. It still doesn’t support all optional arguments but should do the job in most cases. A reference manual is here.
A mini-vignette (more to follow) showing how to run a lookup on a postcode, turn the result into a data frame, and then create an interactive map with leaflet:
The code above produces a data frame with key information
$ postcode EC1Y 8LX
$ quality 1
$ eastings 532544
$ northings 182128
$ country England
$ nhs_ha London
$ longitude -0.09092237
$ latitude 51.52252
$ parliamentary_constituency Islington South and Finsbury
$ european_electoral_region London
$ primary_care_trust Islington
$ region London
$ lsoa Islington 023D
$ msoa Islington 023
$ incode 8LX
$ outcode EC1Y
$ admin_district Islington
$ parish Islington, unparished area
$ admin_county NA
$ admin_ward Bunhill
$ ccg NHS Islington
$ nuts Haringey and Islington
$ admin_district E09000019
$ admin_county E99999999
$ admin_ward E05000367
$ parish E43000209
$ ccg E38000088
$ nuts UKI43
and an interactive map showing geocoded postcode as a blue dot:
One of my latest tasks at work was to analyse data related to Brexit referendum results and the UK housing market.
Luckily, all but rental data (acquired from Zoopla) was publicly available. Property prices and rental prices needed some wrangling as Land Registry doesn’t provide information about Local Authority districts, and that was the unit used by The Electoral Commission. LA districts are not a default geographic category in Tableau (version 9.3.5) but the official blog has recently featured a post demonstrating how to use non-standard mapping.
The final result was a map (below) and a press release. This is another housing market analysis that gained a lot of media coverage, among others by International Business Times, Business Insider, and Mortgage Introducer.
I wanted to dig deeper into the relationship between the voting pattern and the housing market information so I created the following bar charts:
Once the data is visualised in this way it becomes rather obvious that the areas where house prices and the capital gains (yearly average, in the last six years) were the highest, were also the ones that were the most likely to vote remain. The situation is much more difficult to interpret when the the results are sorted by the rental yields. In that case the voting pattern is not that clear anymore.
The scatter plots (and overlapping trend lines) make it easier to see the positive correlation between the percentage of people voting remain and the following variables: median house price (2016), median rental price (2016), and capital gains (yearly, across 2010-2016). This means that as the percentage of remain voters increases, so do the variables mentioned. This relationship did not hold for rental yields where it doesn’t seem to be any relationship between the two.
The Guardian and BBC conducted similar analyses comparing voting patterns to demographic variables.
Kaggle released new data set which I thought would be perfect to try interactive visualizations from qtlcharts, ggvis, and radarchart packages.
The html report generated with RMarkdown and the latest version uploaded to Kaggle (kernel) is here.
Some time ago I discovered London Datastore, a governmental data repository publishing a wide variety of interesting data sets. One of the data sets that drew my attention was describing the composition of the school population in England by first language. Being a non-native English speaker myself, I decided to see whether I could see any interesting patterns and to create a set of choropleth maps.
These maps show that the higher percentage of primary and secondary schools pupils, whose first language is English, tend to occur in the outer London boroughs, e.g. Havering, Bexley, and Bromley. On the other hand, larger percentage of pupils, whose first language is not English, can be found in boroughs in East London (with Tower Hamlets and Newham having especially large percentage).