Q&A Session 2
COMU3120 [Digital Analytics] - Cedric Courtois, PhD and Zoe MacLean, BA
1. Common questions
2. Round of questions
3. APA and reporting
4. Make prettier graphs
5. More questions!
Schedule
The first script on Blackboard doesn’t work…
Yes, that’s because Billboard changed its html page structure a couple of months ago. Not a
problem because the outcome can’t be improved on. Describe as if you did write/run it
Common questions
Explanation of the billboard scraper (Second Module
3 tutorial): https://youtu.be/_SKBGur4Eqg?t=5916
Where can I improve?
The second script (and then running the third script as is)
Common questions
Explanation of the Spotify code (Second Module 4
tutorial): https://youtu.be/Lr79r_hd_Vg?t=5110
I run the second provided code file and get all the track ids. Is that normal? Is it enough?
• Seems to depend on your account/API key how performant the Search endpoint
is. If you experience this too, we ask you to use an API key generated from this
app: https://digitalanalytics.id.au/tokengenerator
• Ensures that everyone gets the same results and that you can improve on the
code to deal with this issue.
Common questions
Anything you’d like to know, clarify, talk about?
Questions right now…
1. Univariate or bivariate?
2. Measurement types (numeric versus categorical)
3. Visualisation – follows the choice of statistics
Pick the right analysis
Difference between numeric and categorical variables (!)
Numeric Categorical
• Values are always numbers
• Makes sense to calculate average
(or other statistics)
• Definitely when values are strings
• Could be numbers, but those are
just labels
• Might be ordered, but the distance
between the numbers is unequal
Examples:
• Age in years
• Distance between UQ and CBD in km
• Number of students in a classroom
• Popularity on a scale 0-100
• …
Examples:
• Gender (even when coded 0,1,2)
• Name (e.g., Kelly, Laura)
• Education level (e.g., primary
education, secondary education,
tertiary education)
• …
• Percentages: always in whole numbers (not 54.45%, but 54%)
• Decimals: two digits after . (e.g., 34.45 instead of 34.452222234, 356.50 instead of 356.50111)
• p-values: three digits after . (e.g., p < .005)
APA-style reporting of numbers/statistics
P-value in output Notation (option a), one of: p > .05, p
< .05, p < 0.005, p < .001 - Preferred
Notation (option b): exact
value
0.0009 p < .001 p = .001
.03 p < .05 p = .03
.004 p < .005 p = .004
.34 p > .05 p = .34
How to report?
The average score for danceability is 0.64 (SD =
0.15) and 0.16 for acousticness (SD = 0.21).
or
Danceability (M = 0.64, SD = 0.15) and
acousticness (M = 0.16, SD = 0.21) were measured
on a scale from 0 to 1.
– Make sure you do not plagiarize!!!
How to report?
The popularity of songs on Spotify differs significantly per decade (χ2(2) =
1106.91, p < .001). More specifically, the popularity tends to increase per
decade: 38% of songs in the 1990s categorise as popular, compared to 57%
of songs in the 2000s, and 80% of songs in the 2010s.
– Make sure you do not plagiarize!!!
How to report?
There is a significant correlation between the popularity
of a song and its danceability (r(8988) = .04, p < .001).
– Make sure you do not plagiarize!!!
- 2
Statistical test of two numeric variables (and how to interpret the
graph): https://youtu.be/QuPq9W0fEX8?t=5624 (* If you missed the
module 5 tutorials, please watch the recordings)
Visualisation aesthetics
• Bar colours
https://replit.com/@CedricCourtois/
FussyFullPrinter
Added (and optional) – Spice up graphs
https://imagecolorpicker.com/en
• Line colors
• Line styles/weight
• Marker styles
https://replit.com/@CedricCourtois/
PowerlessRawMethods
Added (and optional) – Spice up graphs
• Marker colour
• Marker opacity (alpha)
• Red trendline
• Grid
https://replit.com/@CedricCourtois/
VelvetyBusyGraphs
Added (and optional) – Spice up graphs
• Bar colour
• SD/M labels
https://replit.com/@CedricCourtois/
KhakiFrightenedReciprocal
Added (and optional) – Spice up graphs

• Generating figures in Python is ideal to generate standardized reports,
based on frequently updated data and/or quickly create graphs in bulk
through loops (i.e., corporate/policy environment) e.g.,
https://replit.com/@CedricCourtois/VoluminousElatedPrintablecharacter
• Go through the pain of designing nice customized (branded) high-
resolution graphs once, use it again and again by just hitting enter…
• Animated graphs (render multiple graphs, bundle in a GIF)
https://towardsdatascience.com/probably-the-easiest-way-to-animate-
your-python-plots-f5194ebed75f
Just so you know…
digitalanalytics.id.au stays online
(… and will expand/update in the future, e.g., more modules, including customising visualisations)
Generic password ‘guest’ will persist –You can use it later on
Beyond COMU3120 and UQ
Anything you’d like to know, clarify, talk about?
Questions right now…
SECAT surveys are up
Please share your experiences, it’s greatly
valued and appreciated
Feeds into right course development and
adjustments (past and future)
Please evaluate