It is surprisingly fun to play around with the filters and see correlations unfold. If you restrict only a narrow band for number of downloads and leave everything else open, you'll see ratings go up, down, and then up again.

This is really awesome. Great way to get an overview of which apps are popular. Can you make the names of the apps link to the app on the app store?!

The green line showing the average is very helpful to see the distribution in each dimension. A box plot would provide even more information.

The bezier curves look beautiful on the quantized dimensions: ratings, keywords, price.

diziet's example is released under BSD if you want to adapt it to your data. If you're interested in adding parallel coordinates to an existing application or start from scratch, I'm working on a library here:

http://syntagmatic.github.com/parallel-coordinates/

Fantastic visualization - the filters really make it interesting!

Could I suggest an extra column - Downloads * price

For anyone interested - using the console & a quick bit of js the top "profiters" (dowloads * price) were:

    Cut the Rope - $26433990
    Infinity Blade - $15340390
    Ski On Neon - $15246990
    Minecraft – Pocket Edition - $12519090
    Infinity Blade II - $11680290
    Cover Orange - $9712890
    Tap Tap Revenge 4 - $8061490
    Air Penguin - $7762590
    Fruit Ninja - $7633890
    Draw Something - $7268690

        data.forEach(function(d){
           d.profit = (d["Est. Downloads*"] * d.Price);
        })
        data.sort(function(a,b){return b.profit-a.profit});
        for(i = 0; i < 10; i++){
           console.log(data[i].name + " - $" + data[i].profit);
         }

nice! We thought about displaying profit from downloads * price but we found that for most of the big apps, they tweak around prices constantly so the profit would vary too much based on when you poll their prices.

Keep in mind though, like makeshifthoop said, price changes (and even more so, in app-purchases) can have another huge effect on total "profiters" -- some top grossing apps are actually free and rely on in-app

Sorry for coming down harsh but I don't understand how to interpret any of the individual curves, since it looks like they are five points on five different axis with a curve fit through them. What's the significance?

Each curve/line is an individual app with each point representing that app's stats in downloads, reviews, etc. You can use the different axis to see trends. Ie, there is (obviously) a correlation between number of downloads and reviews. You can check out a bit about the history of the visualization at http://en.wikipedia.org/wiki/Parallel_coordinates

If the axes were reorderable, you could move Downloads next to Reviews and see the relation without filtering.

Parallel coordinates is based on projective geometry, and has a point-line duality with orthogonal (cartesian) coordinates. Points in orthogonal coordinates become lines in parallel coordinates.

Few parallel coordinates applications take advantage of the fact that you can show lines in orthogonal coordinates as points in parallel coordinates though. It's one of the few effective ways of representing hyperplanes and hyperlines. Interpreting that is tricky though. Even people who "get" traditional parallel coordinates would have a tough time at first.

Think of it as a multidimensional scatterplot, where lines are a necessary concession to map N-space to 2-space.

http://en.wikipedia.org/wiki/Duality_(projective_geometry)

That makes a lot of sense! thanks for the link :-)

I don't see you as coming down harsh at all, it's a data visualization that requires some math skills. Having an interest in, or understanding of higher dimensional geometry is not for everyone. Basically you can spot positive and negative correlations using these kinds of visualizations.

You could have skipped the first two sentences and got your message across better.

Thats a smart way of implementing a parallel coordinates chart. I like the selection.

Funny tidbit: on firefox especially, forEach and Map functions have a big performance hit compared to doing a native for loop when you're handling a lot of data!

It's really interesting to play with this stuff!

Brilliant!