Analysis package: Podlover

#1

Over the holidays, I’ve taken my initial code to analyze Podlove downloads in R and have written a real R package for it.

Here it is it’s called podlover:

The package can do many more things than the functions before:

  • Fetch necessary tables from the Wordpress MySQL
  • Connect and clean tables for download analysis (either directly from the database or via local CSV files)
  • Plot download curves by episode title, episode number, source, context, client_name, client_type, os_name or any other user-generated variable (plots are available with cumulated or non-cumulated data, with relative or absolute release dates, and with all kinds of graphical output options)
  • Generate average performance tables per episode for total and average downloads as well as average downloads for specific periods (e.g. the first three days after launch)
  • Analyze performance of launch against post-launch in a four-box grid
  • Create regression models of downloads against episodes or episode launch dates at a specific point in time (similar to @rahra 's approach), both as statistical models with coefficients and graphically
  • Create example podcast data to play around with the package

All functions are documented in help files. The README.md on GitHub (=package vignette) shows many examples and details. Test, feedbacks and contributions are welcome!

4 Likes
#2

Looks great - can your add-on work with show module podcasts?

#3

Without knowing how shows are represented in the data, my guess is “not yet”. What I would need to know to estimate the effort is:

  • Is there a fundamental difference in the data structure as opposed to single podcasts?
  • Does use of the show module create new tables or change table definitions?
  • What fields in which table define what show an episode belongs to?
#4

The data structure is not different - I think there is just one more filter needed, where the podlove post id needs to be sorted a long the show module.

Shows module adds tables to term_taxonomy and terms and termmeta.

objectid in term_relationships references to the id of an episode in podlove_episode

#5

So if I understand correctly, shows are related to posts in wp_posts. Is there a show identifier in wp_posts itself or does the post_id relate to another table containing the show label?

It’s a little more work than just the filter - I assume I will have to fetch an additional table to identify the show in a userfriendly way.

#6

no objectid of terms_relationships references to the episode table of podlove of postid.

#7

I’ve published V.1.3.0 of podlover, which includes:

  • Add summary function option to podlove_plot_curves(): This allows you to add an optional curve (e.g. median, mean, max…) to a multi-curve plot, giving a reference to the curves. This is especially useful in cominbation with the last_n option, as the summary curve will remain based on the whole set of downloads, not just the ones plotted.
  • Add a limit function to podlove_plot_curves(): In combination with the last_n and summary curve option, this allows you to see the either the full picture or just the ones of the last plotted curves.
  • Add a legend option to podlove_plot_curves(). This more classical way to show which curve is which can be used instead of labels inside the plot. Thanks to @PechGehabt for the idea!
  • Updates to unit tests, vignette and readme
  • Add new time scaling options to podlove_plot_curves(): You can now use hourly, daily, weekly, monthly and yearly aggregations.
  • Add new options for performance labelling: Allow choice of performance plot labelling variable and introduce a add legend option (displays a reference table)

Bugfixes

  • Enlarge plotting area in curve plots to ensure labels aren’t cut off
  • Reduction of unnecessary messages within functions

The new version can be installed as usual via devtools/GitHub:

# install devtools if you don't have it already
install.packages("devtools")

# install podlover from GitHub
devtools::install_github("lordyo/podlover")

Examples:
Multicurve with legend and aggregation function:
plotcurves

Performance Plot with new labels and legend:
podlover_performance

#8

Looks great - have you had the time to have a look at the show module?

#9

I have received some sample data from a multi show podcast. I hope I can put that to good use.