Diverge - Emerge 2.0

In the exploratory process to create Data visualizations, it's important to generate a process of divergence-emergence and convergence of the different possible designs we explore. Therefore, in addition to the sketches presented the last week for the Google Playstore dataset, we present today a total of 29 sketches which were analyzed to get to the third step of the selection process: convergence.

With these 29 sketches as starting point, we decided to narrow the list considering diferent elements. In one hand the Question(s) we want to answer and in other hand the range of possible relations that my dataset allows me to explore. What kind of app do I want to develop? should be the initial question of a developer for Android apps, and in that direction, the data we have could help in order to focus on some features. What is the ideal size of the app? The most popular category? The most/least atomized category?.

The sketch FG12 relates the Android version to the amount of downloads (both categorical dimensions) and in each subsection we ca explore the quantity of apps and how they relate in Size and rating. Given that the categories are too many to add in this graph, if we want more information for some specific subsection (For example, if we aim to develop an app which gets to +100,000 downloads) we could get an specific detailed view of this segment with a FG10 type of chart. With respect to what my data allows me to develop, some charts were discarded due to the absence of numeric data to apply them or due to the low correlation between some of our variables.

Lastly, even if we could still modify the model we are choosing, we already started to work in the implementation of this sketch in p5.js. This is the next step in our process to visualize the Google Playstore dataset and a preview to this task is included in the image below.

This is a first translation of the sketch 12 idea to p5.js

Sketch 1: Number of Apps for each Android Version ordered by Rating or Size. Every color represents a category, and we could interactively change between Size or Rating on the y-axis.

Sketch 2: Chart made of Android figures representing the Top Categories of Apps. The amount of installs is represented by the size of the figures and its average rating by hue (color).

Sketch 3: Hive Plot of Android Versions. Each point represents an App which is available from 1.0 until the last point it has for the other versions. Color represents categories.

Sketch 4: Sorted stream grapgh. Each stream is a category where the upper line follows the path of rating and the height of it the average Size of the app.

Sketch 5: Plot with Clusters of Apps by Category, linking every app to its date of update. The size of each buble represents the file size.

Sketch 6: Branches' Visualization, representing a version of Android in every main branch and its 'sub-versions' or dependants coming from it. The line graphs and bar charts add detail to the timeline and to the composition of each version (height represents the amount of apps for that version).

Sketch 7: This sketch combines what I also think for others. Here I mix the information of Number of downloads, and for each 'bin' I explore its size and last update distribution as well as other features (Type, Content, Android version,etc.)

Sketch 8: This is a simpler implementation of Apps Clusters by amount of installs, where the size of the bubbles represents the size or rating of an app.

Sketch 9: This wheel plot is intended to join several pie charts as apps by rating, number of installs, Type, content, etc.

Sketch 10: Based on the billion-dolar-o-gram this chart shows the distribution of categories (by number or percentage). This could be used as a detailed view for a cluster.

Sketch 11: Android Version by amount of apps available and its composition by downloads for each version.

Sketch 12: Table with Installs in the x-axis and android version in the y-axis, what allos us to have a detailed view for each segment, where we can compare size vs rating with the marginals in the bottom ant to the right of the table.

Sketch 13: Heat map for rating and amount of downloads with the marginal probabilities in the margins.

Sketch 14: Distribution by size for each rating level, considering a previous grouping by Android version (It could also be by Installs).

Sketch 15: Raw idea of a Scatterplot for variables rating and size, including the maginal distributions on the margins.

Sketch 16: Raw sketch relating Downloads and Ratings by size.

Sketch 17: Violin plots by Installs (more effective than Version) with Rating on the y-avis.

Sketch 18: Boxplot for every Android version (Number of Installs would be an interesting variable to inspect here) and ratings.

Sketch 19: Round Plot with coloured lines for Rating, Size, and Installs. Points around the circumference represent Android verson (or amount of installs).

Sketch 20:Variation of a bloxplot where the width of the blocks illustrates the number of apps for each Android version.

Sketch 21: Radviz plot for the numerical variables in the dataset.

Sketch 22: Evolution of Categories depending the Android version in rating and installs.

Sketch 23: 3D plot which includes bars to see the distribution by category for teh variables size, rating and installs.

Sketch 24: Scatterplot with an added grid to explore the relation between rating and reviews by segments.

Sketch 25: The size of Bubbles indicates the amount of apps in every Installs category and its color (hue) the average rating.

Sketch 26: Distribution by size of the apps, highlighting some points of interest with labels showing more specific information.

Sketch 27: This one mixes ehat I included in other sketches. The 'dolar-o-gram' plot is now displayed for every Andoid version where we can also add when was every app updated (color lines indicating date and category).

Sketch 28: This was a simpler and preeliminar version of sketch 27.

Sketch 29: This was also a preeliminar idea to sketch 27, where the line illustrates Android versions (time) and eachcolor line is an app updated in a certain date.