/sci/ - Science & Math » Thread #14184108

70KiB, 1820x666, data.png

View Same Google iqdb SauceNAO

Making a Grid-like Figure in R (or Python)

Anonymous Fri 04 Feb 00:54:04 2022 No.14184108 View Reply Original Report

Quoted By:

I've kept data on myself since January 01, 2020, so about 765 days of data. See pic rel for the top of the data.
One variable is the type of exercise I have done on each day.
I want to make a figure that indicates the type of exercise I did on each day.
Like, a 109 x 7 grid, where each cell is colored according to the (factor / string) variable that indicates what exercise I did on that day.
I'm having a hard time coding this and am wondering if anyone has any suggestions.
If you are curious, in the past 765 days I have consumed 999 alcoholic drinks, 826 "units" of weed (a unit is approximately equal to one and a half typical hits), and 2337 cigarettes.

Anonymous

Anonymous Fri 04 Feb 2022 00:54:59 No.14184110 Report

Quoted By: >>14184198

Oh, and I'd like to do this in R, but I guess I could also do it in Python. I would prefer R by a fair margin, though.

Anonymous

Anonymous Fri 04 Feb 2022 00:58:00 No.14184119 Report

Quoted By:

Oh wait, I'm onto something... Will post solution if I get it.

Anonymous

Anonymous Fri 04 Feb 2022 01:32:38 No.14184198 Report

Quoted By: >>14184212

>>14184110
You need to get your data in to 'tidy' format.

library(tidyverse)
yourData %>%
gather(-dayDate, -dayNum, -dayWeek, -location, -exType, key = activity, value = value) -> longData

If you want a heatmap, use geom_tile.

longData %>%
ggplot(aes(x = dayNum, y = activity, fill = value)) +
geom_tile()

Alternatively, you could facet by exercise type.

longData %>%
ggplot(aes(x = dayNum, y = value) +
geom_line() +
facet_wrap(~activity)

Anonymous

View Same Google iqdb SauceNAO Rplot.png, 7KiB, 500x1000

Anonymous Fri 04 Feb 2022 01:37:53 No.14184212 Report

Quoted By: >>14184228

>>14184198
Nice, appreciate it.
I'm getting close, also using geom_tile.
The last piece I'm stuck on right now is getting the colors right (specifically, getting "None" to be white).

I did some light manipulation to prepare it for the grid, but generally my data were in good shape (I pride myself on that).

dat$dayOfWeekNum <- ifelse(dat$dayWeek == "Sun", 0,
ifelse(dat$dayWeek == "Mon", 1,
ifelse(dat$dayWeek == "Tue", 2,
ifelse(dat$dayWeek == "Wed", 3,
ifelse(dat$dayWeek == "Thu", 4,
ifelse(dat$dayWeek == "Fri", 5,
ifelse(dat$dayWeek == "Sat", 6, NA)))))))

dat$weekNum <- (dat$dayNum + 2) %/% 7

(I made exTypeNew where I bundled some of the less common exercise types.)

ggplot(dat, aes(x = dayOfWeekNum, y = -weekNum, fill = exTypeNew)) +
geom_tile() +
theme_classic() +
theme(axis.text=element_blank(),
axis.ticks=element_blank(),
axis.title = element_blank(),
axis.line = element_blank()
) +
scale_fill_manual(values=group.colors)

Anonymous

Anonymous Fri 04 Feb 2022 01:38:55 No.14184215 Report

Quoted By:

Take a look at the package kableExtra. With Rmarkdown you can render neat otputs in .htm or .pdf files.

Anonymous

Anonymous Fri 04 Feb 2022 01:40:51 No.14184221 Report

Quoted By: >>14184237

I'd also like to make each cell a square rather than rectangular, that would be a second piece.
Also, ideally, I would have the pieces not touching (i.e., small gaps of white separating them), but that might be more effort than I'd like to put forth.

Anonymous

Anonymous Fri 04 Feb 2022 01:44:06 No.14184228 Report

Quoted By: >>14184236

>>14184212
Tidy format is in 'long' format. Yours is in 'wide' format. Every exercise needs its own row for each day. In your visualization, you can only show one exercise per day. You also devote two dimensions to plot time when you could plot it all on one axis, freeing up another dimension. What happens if you biked and played basketball on the same day? By plotting exercise on the y axis, you can use color denote the intensity or duration of the exercise.

Cool dataset anon.

Anonymous

Anonymous Fri 04 Feb 2022 01:50:59 No.14184236 Report

Quoted By: >>14184245

>>14184228
I consider it in long format, there is only one row per day, and I only allow one exercise per day (I typically only do one form of exercise per day - I included some hard coding for days where that wasn't the case, which were few).
I could much more easily plot it in one dimension, but I find that less easy to digest - I like the N x 7 grid (7 b/c 7 days in the week and I follow typical patterns; e.g., working out less on Sundays).
The other exercise-related variables I have are run times and some distance related data (for bike and run). I could include those in future plots, which would be interesting.
But, for now, I like this Nx7, one exercise per day plot, and I'm trying to get it perfect.
Another future project is to use this data to create a predictive model of my run times (n = 100 runs), which is pretty good, and I have the distance and will make lag variables for my substance abuse (e.g., number of cigarettes smoked in past 7 days, with more recent days weighted more heavily).
Appreciate the comment and compliment, apologies if I sound ungrateful or snarky.

Anonymous

Anonymous Fri 04 Feb 2022 01:51:00 No.14184237 Report

Quoted By:

>>14184221
So try the GGcal package
https://karobben.github.io/2020/06/19/R/GGcal/

Anonymous

Anonymous Fri 04 Feb 2022 01:55:48 No.14184245 Report

Quoted By: >>14184262

>>14184236
No worries. I may be misunderstanding how you're encoding your observations. Also, to get that border around your tiles:

ggplot(dat, aes(x = dayOfWeekNum, y = -weekNum, fill = exTypeNew)) +
geom_tile(color = 'white') +
theme_classic() +
theme(axis.text=element_blank(),
axis.ticks=element_blank(),
axis.title = element_blank(),
axis.line = element_blank()
) +
scale_fill_manual(values=group.colors)

Anonymous

View Same Google iqdb SauceNAO Rplot2.png, 6KiB, 400x1200

Anonymous Fri 04 Feb 2022 02:02:25 No.14184262 Report

Quoted By: >>14184285

>>14184245
awesome, didn't know about
>geom_tile(color = 'white')

here she is, i think it looks pretty good!
of course, could keep futzing with the colors, and i may, but the gist is certainly there. a nice fun thursday night project

final ggplot code:

ggplot(dat, aes(x = dayOfWeekNum, y = -weekNum, fill = exTypeNew)) +
geom_tile(color = 'white') +
theme_classic() +
theme(axis.text=element_blank(),
axis.ticks=element_blank(),
axis.title = element_blank(),
axis.line = element_blank(),
legend.title = element_blank()
) +
scale_fill_manual(
values = pal,
limits = names(pal)
)

Anonymous

Anonymous Fri 04 Feb 2022 02:11:27 No.14184285 Report

Quoted By: >>14184306

>>14184262
I recommend using premade palettes or color palette generation tools. People are generally bad at doing it on their own. I taught a data viz class to some grad students last year. That was one of the biggest points I tried to get across. Scientific literature is riddled with hideous and downright bad color selection. Here's a neat tool that will generate a custom palette from an image.

https://github.com/leonawicz/imgpalr

Anonymous

Anonymous Fri 04 Feb 2022 02:19:29 No.14184306 Report

Quoted By:

>>14184285
i emphatically agree and empathize. i tried to use colorbrewer but was frustrated because the auto palettes it generates were limited to 9 levels and I had 10 levels. I suppose I could have made the 9 level one and then added a tenth... maybe.
will check out that package, appreciate your sharing it.
to note, i tried to do the colors in a way that was somewhat meaningful to me (e.g., the XC is light blue for snow, run is black for pavement, hike is green, etc.). but yeah, ugly no doubt.

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches