## Friday, 8 June 2012

### Homage to Minard

It is amazing to me that I can work with a product like QlikView for 6 years and still come across new things all the time.

Recently, my good friends The QlikMonkey, posted a visualization about why the soccer team, Manchester United, lost the English Premier League:

They were also kind enough to share some other work that they had done to create similar parallel coordinate charts like this.

One of the things that I noticed about it was that some of the lines had different widths - I hadn't known that we could do this in QlikView!!!

I had always known about the different options that we could add expressions to under an expression in a chart, e.g. Background, Bar Offset, Text Color, etc.  One of those is "Line Style".  I had thought (and this is correct) that was to do with creating different types of line such as dashed.  I didn't realize that we could also use this property to dynamically change the widths of lines.

This is from the QlikView Help file:

"The relative width of the line can be controlled by including a tag <Wn> where n is a multiplying factor to be applied on the default line width of the chart. The number n must be a real number between 0.5 and 8. Example: <W2.5>."

It occurred to me that I might be able to use this feature to replicate one of the classic charts in visualization history - Charles Minard's chart of the losses suffered by Napoleon during the disastrous Russian campaign of 1812/13, which he drew in 1869.  Even Edward Tufte, one of the modern-day guru's of data visualization, described this as "probably the best statistical graphic ever drawn".

I created a data set of X, Y coordinates along with the volume of soldiers for each point (field called "Count").  A separate set for each direction (and spur line).  I then created the line chart and used a % of the Max value to calculate the appropriate "W" value for the line:

='<W' & Round(0.5+(Sum(Count)/Max(total Aggr(Sum(Count), X))*7.5),0.1) & '>'

And here are the results:

Obviously this is not an exact replica.  If I was doing it again, I would include more data points to get finer detail.  But as a proof-of-concept, I believe it is a good enough effort.

Stephen Redmond is CTO of CapricornVentis a QlikView Elite Partner