topic A Challenge - just for fun in QlikView

A Challenge - just for fun

stephencredmond — Tue, 14 Jul 2009 23:55:07 GMT

See http://community.qlik.com/media/p/70295.aspx for the QVW

In the following BBC article: http://news.bbc.co.uk/2/hi/uk_news/magazine/7937382.stm, the chart below was touted as being an excellent way of visualising data. I think that it is crap.

Propose a better way using QlikView objects.

Stephen

A Challenge - just for fun

Wed, 15 Jul 2009 00:30:32 GMT

The advantage of the original graph is that it shows the relative size of 3 different dimensions. It's impossible from my submission to tell the relative size of crew versus passengers. But I think my submission tells a more compelling story at first glance. It was good to be a rich woman on the Titanic.

A Challenge - just for fun

Anonymous — Wed, 15 Jul 2009 08:50:15 GMT

Without actual trying, I'd think that the 3-dimensional grid chart would do the job. Dimensions: group (class, crew), sex, and "survived", expression is count.

A Challenge - just for fun

bismart — Wed, 15 Jul 2009 18:34:55 GMT

A Challenge - just for fun

Wed, 15 Jul 2009 18:54:07 GMT

This one is using a trellis chart...

A Challenge - just for fun

Wed, 15 Jul 2009 18:56:32 GMT

Hi,

I think the chart is good...just that it's not-so-known and bit difficult to understand. Trelis can achieve this to an extent but some questions remain unanswered:

1) Overall data distribution for a dimension (e.g. male/female ratio). Looking at the new chart we can easily depict the female/male ratio for each class.

2) We can drill down to more detail by adding more dimensions. In the titanic example, we can add fourth dimension 'Region' and can answer questions like how many surviving males travelling 1st class were from Europe?

And last, all this is one single chart.....it would be great if we can do this in QV.

Thanks

Amit

A Challenge - just for fun

Anonymous — Wed, 15 Jul 2009 19:09:36 GMT

Bismart was very close to my suggestion, just a wrong (in opinion) style. This is what I meant
It is better than trellis because it shows the relative numbers visually.

A Challenge - just for fun

stephencredmond — Wed, 15 Jul 2009 19:16:02 GMT

Hi Amit,

One of my main problems with the chart presented is the arbitrary order of the dimensions. If you change the order of any of the dimensions, the shape of your graph is completely different.

Also, the overlapping colours (again, the colours are arbitrary and mean nothing) is potentially confusing.

I would love to hear Stephen Few's opinion on this one. Maybe I will email him about it.

Regards,

Stephen

A Challenge - just for fun

Wed, 15 Jul 2009 19:32:48 GMT

My first try was to have the same approach as Michael; a grid chart with "pie bubbles". The only drawback is that you can't compare the "yes/no" rate for female crew. Another problem is that it's not really utilizing the space very well. But I still think it beats the original visualization! 🙂

A Challenge - just for fun

Anonymous — Wed, 15 Jul 2009 19:39:17 GMT

kjn,
The number of female crew is very small, hence that little dot, but you get a nice 50/50 pie if click on it.

A Challenge - just for fun

stephencredmond — Thu, 16 Jul 2009 00:58:12 GMT

Hi All,

Stephen Few's response:

From: Stephen Few
Date: Wed, 15 Jul 2009 09:43:51 -0700
To: <Stephen.Redmond@capventis.ie>
Subject: Re: New chart type

Hi Stephen,

Unfortunately, in the BBC article Robert Kosara's chart was displayed out of context. It actually solves particular problems associated with multivariate analysis by improving a type of visualization called parallel coordinates, which can be quite powerful, but look ridiculous upon first encounter. To put Kosara's graph into context, you might find it helpful to read my article about parallel coordinates and then to read the research paper that Kosara and his students wrote regarding parallel sets.

I hope you find this useful.

Steve

A Challenge - just for fun

Thu, 16 Jul 2009 01:57:24 GMT


Stephen Redmond wrote:
 I think that it is crap. <div></div>

+1000. There are some pretty terrible visualizations out there. Here's one of my favorites from the Onion:

A Challenge - just for fun

johnw — Thu, 16 Jul 2009 04:52:59 GMT

I think the main problem with the original is that it is a very unfamiliar type of chart to most of us. It took me quite a while to make sense of it, but the meaning of it now stands out clearly to me when I look at it.

I'm not usually a fan of pie charts, but in this case, I do like Michael's Pie charts. I'd personally remove a lot of the "noise" - lines, borders, background color, shiny effects on the pie, legends, and so on. I'd also use a pale red and green for our traditional associations with good and bad, with enough contrast between the two for people with color blindness to still make sense of it.

A problem with the pie charts in this case is that while you get a good idea, say, that there were a lot more crew than 3rd class passengers, it's really hard to tell how many more since we're being asked to mentally add up two different areas of two pie shapes, and then compare the two total areas. That's simply not something the human mind does well. You can see, for instance, how much simpler this task is in the original chart since these totals are indicated by much more easily-compared line lengths.

So to go heavily against the flow here, I feel that the first chart actually presents the most information the most clearly, and allows for the greatest understanding of the numbers. Other charts that have been presented do a better job of demonstrating certain aspects of the numbers, but I don't feel that any of them have better presented the entire picture.

One reason we may be looking at it with such disfavor is that we're so used to being able to do further analysis. Want to see that pie chart of female crew that is just a dot in Michael's chart? Just click on it. Want to combine male and female passengers for an analysis? Just remove it as a dimension. But I assume the context of the original is that it is a single, static chart intended to convey the most information as clearly as possible. In that, I personally feel that it is a great success.

(Edit: I'd further simplify the original chart by removing the legends. There's no need to say "Class" or "Sex" when it is obvious from the values, and the only reason "Survived" is necessary is because it was put in terms of Yes/No instead of Lived/Died. I'd also remove the black border. And the little grey bar below each legend on the left is complete fluff and should be removed.)

A Challenge - just for fun

Karl_Humma — Thu, 16 Jul 2009 05:03:06 GMT

Looks like they were trying to mimic the French Train Schedule from the 1880s in Edward Tufte's book The Visual Display of Quantitative Information. An interesting blog about it can be found here.

http://marlenacompton.com/?p=103

Good book btw...

A Challenge - just for fun

Anonymous — Thu, 16 Jul 2009 06:38:51 GMT

John, The original chart does present a lot of info. I feel it like a graphic interpretantion of a pivot table, a sort of. Wouldn't mind to have it in QV one day. (As for my "fancy" example - agree it could be clearer. I just took the defaults as is, excpet a couple minor things.)

A Challenge - just for fun

rwunderlich — Thu, 16 Jul 2009 12:56:23 GMT

How about the venerable horizontal bar chart?

Re. :Re: A Challenge - just for fun

Thu, 16 Jul 2009 14:35:12 GMT

Not completely surprised that Horizontal Bar Chart is the better...

Stephen Few will be totally agree with your smart analysis, Rob ! Nice Shot !

A Challenge - just for fun

johnw — Fri, 17 Jul 2009 00:18:18 GMT


Rob Wunderlich wrote:How about the venerable horizontal bar chart?

Reasonable at displaying some of the facts, and probably better than the multiple pie charts. But what percentage of passengers were female? What percentage of male passengers survived? What percentage of female passengers were in first class? Your bar chart mostly gives you information at the lowest level of detail. Summary information requires you to mentally add together the lengths of bars or bar segments. In the original, these summary facts are easily seen once you figure out how to read the chart.

A Challenge - just for fun

Fri, 17 Jul 2009 00:51:20 GMT

It seems to me that we've hit the QlikView charting wall. To me, this has become QlikView's weakest area. I can't build the original chart, but I still should be able to get the same answers without starting a Forum thread!

Beginning with a bar chart, the user should be able to interact with it how they want. A pivot table allows the user to swap column order or collapse lower-level aggregation, but charts do not. And chart formatting has become a tedious task. I'm sure many of the users in this thread have used the nearly secret Ctrl-Shift method to format chart objects.

I've been drafting a blog post on this. I believe QlikView should overhaul charting so that users can intuitively add/remove fields, reorient graphs, convert to trellis, and more, while automatically formatting in an intelligent and attractive way.

A Challenge - just for fun

Fri, 17 Jul 2009 01:09:04 GMT

And let me head off the "end users can create their own objects" thread. I don't believe the chart creation interface should be forced on a typical end-user. It is cumbersome and confusing. It is filled with options that are respected by one chart and ignored by another, and even by different presentation styles of the same chart!