Wow, Thomas, thank you so much for your always kind comments. I really appreciate it.
1 Like
Good work, I liked the idea of correlation you chose; it sounds interesting. Another suggestion: you could also try logistic regression to understand which of these characteristics is most relevant when choosing a bet type.
And another one to add to the mix. I know the idea of all things data is not to think about a visual to use the data for but that’s exactly what I did this afternoon.
I wanted to create a mirrored bar chart and the male/female column was a nice start.
What I did was counting the number of yes answers for these habits:
[‘Cigarettes’, ‘Alcohol’, ‘Gamble’, ‘Skydiving’, ‘Exceed speed limit’,
‘Cheated significant other’, ‘Steak’],
That’s what you see in the barcharts, the darker the color the more yes answers
for a person.
Surprise, nobody answered 7 times yes.
You can filter down on habits, if you select only one, say “Cigarettes”, you can see which part
of the total smokes.
There is a filter to change the view (income, age, education). The categorical for ordering, I just saw it in @mike_finko s code, I forgot, but then I remembered that horizontal + categorical is sometimes a warzone. I like to forget my warzones 
Percentage of total in group is in the hover, added later, I missed it.
I think in hindsight, this visual is not a good idea if all habits are selected.
5 Likes
Yes, that’s a good idea - summarize inside cards at the top - either so people don’t miss it or to make it stand out more. I’ll have to try that soon.
Thanks for the feedback!
Good to be back and glad to see Figure Fridays are still going strong! I have a lot of catching up to do 
The reference lines correspond to the values in the bottom “All” bar and were a visual aid to show the breakdown across the entire sample population along that one dimension. For the example you provided, you can see respondents 60+ years old tend to like their steak prepared Rare & Medium Rare more than the average person surveyed.
Another example is smoking. Overall, only 16% of those surveyed smoke but, within that group, you can see the distribution skews younger with 26% and 19% for 18-29 yo and 30-44 yo, respectively.
Another trend I noticed was the lottery question broken down by income. Not surprising, those with the highest household income were most willing to take the riskier Lottery A, while those at the lowest income bracket tended to pick the safer bet.
And thanks for the kind words. There’s a Plotly Discord server?
2 Likes
Thanks for the explanation. Yes, here’s the Plotly Discord server link I forgot to add.
Hi Marianne,
Here I’m sharing the information on the clusters, or part of it, that you recommended, along with the percentages of dominant behavior and typical demographic profiles (not the only one). This gives you an idea. It’s important to note here that these clusters aren’t statistically perfect.
===== The Realistic Moderates =====
Cluster Size: 385 individuos (71.2% del total)
Dominant Behaviour:
- Lottery Choice: Lottery B (51.3%)
- Smoke Cigarettes: No (97.7%)
- Drink Alcohol: Yes (77.7%)
- Gamble: No (53.7%)
- Skydiving: No (100.0%)
- Speed Limit: Yes (100.0%)
- Cheating: No (83.9%)
- Eat Steak: Yes (78.9%)
Typical Demographic Profile:
- Gender: Female (50.1%)
- Age: > 60 (26.8%)
- Income: $50,000 - $99,999 (32.2%)
- Education: Bachelor degree (32.5%)
- Region: South Atlantic (17.7%)
- Steak Preparation: Medium rare (29.4%)
===== The Principled Cautios=====
Cluster Size: 59 individuos (10.9% del total)
Dominant Behaviour:
- Lottery Choice: Lottery A (50.0%)
- Smoke Cigarettes: No (77.2%)
- Drink Alcohol: Yes (50.8%)
- Gamble: No (67.2%)
- Skydiving: No (94.9%)
- Speed Limit: No (98.3%)
- Cheating: No (81.0%)
- Eat Steak: Yes (67.2%)
Typical Demographic Profile:
- Gender: Female (55.9%)
- Age: > 60 (28.8%)
- Income: $50,000 - $99,999 (27.1%)
- Education: Some college or Associate degree (32.2%)
- Region: Middle Atlantic (23.7%)
- Steak Preparation: No especificado (32.2%)
===== Selective Adventurers =====
Cluster Size: 36 individuos (6.7% del total)
Dominant Behaviours:
- Lottery Choice: Lottery B (63.9%)
- Smoke Cigarettes: No (91.7%)
- Drink Alcohol: Yes (83.3%)
- Gamble: No (55.6%)
- Skydiving: Yes (100.0%)
- Speed Limit: Yes (97.2%)
- Cheating: No (77.8%)
- Eat Steak: Yes (75.0%)
Typical Demographic Profile:
- Gender: Female (50.0%)
- Age: 30-44 (41.7%)
- Income: $50,000 - $99,999 (30.6%)
- Education: Some college or Associate degree (38.9%)
- Region: West South Central (16.7%)
- Steak Preparation: Medium rare (27.8%)
===== Risk Hedonists=====
Cluster Size: 61 individuos (11.3% del total)
Dominant Behaviours:
- Lottery Choice: Lottery A (59.0%)
- Smoke Cigarettes: Yes (100.0%)
- Drink Alcohol: Yes (93.4%)
- Gamble: Yes (73.8%)
- Skydiving: No (100.0%)
- Speed Limit: Yes (100.0%)
- Cheating: No (82.0%)
- Eat Steak: Yes (100.0%)
Typical Demographic Profile:
- Gender: Male (55.7%)
- Age: 18-29 (36.1%)
- Income: $50,000 - $99,999 (34.4%)
- Education: Some college or Associate degree (32.8%)
- Region: Pacific (27.9%)
- Steak Preparation: Medium rare (41.0%)
1 Like
beautiful color theme here!
I was thinking the same.Self-selection and nonresponse bias seem especially relevant here. Sometimes the patterns in who doesn’t participate are just as revealing as the data itself.
I’m not 100% sure if I understand your comment, is it about the counting yesses? I had actually more fun to find my profile and the profile of a friend of mine. Opposites 
. The colours btw are just 2 colours with grades of transparancy.