Dataset Browser in Shiny:)

During the early stages of exploring a dataset, the ggpairs() plot is intriguing, and a clever design...

... but it's slow, and depending how much stuff you are trying to look at, the cells are too small to read.

 
# this is the first of two reproducible R sections in this presentation
# you have to load GGally to use the ggpairs() plot
library(GGally)
library(datasets)
ggpairs(swiss)

       plot of chunk unnamed-chunk-1

What about corrgram()?

Another way to see a lot of variable relationships and a lot of correlations in a compact space is Kevin Wright's corrgram.
 
# this is the second reproducible section - the vote dataset is part of corrgram
library("corrgram")
corrgram(vote, order=TRUE, upper.panel=panel.cor, main="vote")
       
It's good, but also a little cumbersome and corrgram can be really slow. There's also nothing wrong with just firing plot() from the console to see things one by one, but what about a sort of dataset browser application? Like maybe written in Shiny?
Why yes, I'm glad you asked...
The purpose of this application is to allow exploring datasets by plotting numeric variables on the x and y axes.

Put things in, take them out, look at the correlations change. It could be considered as a complement to things like ggpairs and corrgram.

You can either work on 'swiss', which is loaded up by default, or load your own csv. The column headers that constitute the fields available to plot will be read in dynamically!
I could have used this thing during the Brian classes..

Just to take one example of how this application might be used, here's a fragment of the report I did for the multivariate-regression assignment in the Regressions class:







There were a lot of variable-pairs to consider. And, in the interest of parsimony, I needed to find the ones that were highly correlated amongst themselves in order to leave some out and avoid variance inflation. The question of working method is partially a matter of taste, but I probably could have saved some time if I could have discovered these things using a Shiny program.

Exploratory analysis by visceral 'feel'
Thank you for your attention - here are the relevant links:
(1) Application
(2) Presentation
(3) R code. "Post your code on github." I put the server and UI portions in one file.
(4) Documentation: simple documentation is given on the application itself