Infographics & Data Visualisation

I'm a sucker for nice diagrams and ways of data visualisations. I read Wired a lot, which is probably where this has come from tbh. Here's one of my favourites:

https://ia.net/wp-content/uploads/2009/07/wtm4-final.png

[The state of the web mapped onto Tokyo's metro system]

Bump - would be keen for someone with a bit more data presentation experience to offer some thoughts on something I'm throwing together. @Chalfie?

Happy to have a look, remembering I'm not that well qualified but I do have an opinion...

Thanks, man. Qualifications can suck it.

Two parts, one super simple, the other very broad - here's the gist:

I've done some statistical analysis of the relationship between words/concepts in texts.

1) I measurements for 5 words over 10 decades (measured against a constant variable). The simplest thing to do would be a line graph with different coloured lines representing each word/concept. HOW DO I DO THIS??!

2) I'm trying to think of interesting visualizations full stop that could be relevant. Not necessarily with regard to the five concepts, just general ideas.. Like, maybe an endogram of some sort. Again, not sure what would work, or how I would do it?!

I have a book called 'information is beautiful', it has loads of random data in graphic form.

if you want inspiration it would be worth checking out

^^ FYI - figured out the line graph bit with excel, I think.

^ cheers, I'll have a google in a bit.

Edward Tufte is generally considered the highest authority on information design. His book Envisioning Information is what you want to read (along with the rest of his writing). David McCandless who writes Information is Beautiful creates the sort of popular and fun types of info graphics that often don't need the complexity of attempting to be an info graphic at all. However, might be a good starting point for exploring different options/inspiration.

Sometimes the best way to start these things off is using the graph tool in Illustrator, inputting your data and seeing what comes out the other end. Then use that to inform how you want to explain the patterns you have discovered... Is it's really data heavy and you want some automation I have no idea where to start with that.

didn't know illustrator had a graph tool. helpful. cheers!

What this guy said ^^

I use one tool and batter it all the time. A panel chart. A line chart with multiple lines is tricky to follow.

Everyone who's anyone with say tufted and mcandless. Personally I prefer Nathan yau and Stephen few. Few is more business stuff, joeslain reckons it all looks like excel chars (which it does to be fair) but it's clear.

If you want, I'm on leave for 5 months and have a couple of books you could borrow?

Bump again. Anyone use ggplot in R? @Chalfie?

I've used it. What are you up too?

I'm trying to make a histogram or density chart in which the Y axis is a value, and the X axis is a feature/string/word AND the width of that feature/string/words representation is equal to a value.

i.e., the individual features are represented in two ways - the total percentage of the population, and the percentage of something else in width.

Does that make sense? Visualizations are not my thing. Even trying to devise ways of visualizing my data is doing my head in.

Does that make sense?

I'm doing a ggplot tutorial right now so may get there eventually..

I'm only just plodding in and around R.
It sounds like you need a histogram (because width of the bin is a thing as well as the height of the bar).

http://www.cookbook-r.com/Graphs/Plotting_distributions_(ggplot2)/

Sorry if I'm not saying anything great here.

What about geom_point where the point value up the y axis is population, but the size of the point is drawn from the 'something else' p + geom_point(aes(size = something else))

I don't think bin width in the histogram would do what you want it to do - it divides your complete dataset into a range of values - these would usually be evenly spaced

I'm not quite sure of anything, I've got the greatest showman soundtrack blaring in one ear and the need to get other stuff ready for today.

So I've been trying with histogram as I had/have the same hunch as Chalfie.

I can split the bins into different widths with breaks, and using a script which takes the width from the dataframe (loop to create a list of entries that look like: 0, var[i-1] + var[i], etc.).

But I can't tell a) if this is actually the correct data (using fill to choose certain "bars" doesn't seem to be working/selecting the correct bars).

My mind is mush.

Loop?
Isn't everything a list in r?

^ the one thing I remember

Do you have one y variable and 2 X? (One is a width and one is a category)

Instead of playing with your real data, could you make some dummy data to play with?

(stack overflow chat)

Only able to help tonight tmrw night if at all. But getting this down might help someone help you

I realize I've been trying to visualize this wrong. Fuck me. Let me restructure my stuff and see if I can solve this.