TalkingViz: Language matters! Bilingual analysis of early school leaving in EU.

This my second post in a series that looks at my Tableau Public visualizations so far.

As you know, I love Tableau and what we the users can achieve in mere minutes, at the same time I like learning new languages. So I was pretty pleased with the announcement that Tableau Communities was going to provide translation on the fly.

#ISpeakData—How about You?

Why does this matter?

Some of the more seasoned forum helpers have already been using Google Translate to help users in other parts of the world where English is not their first language. Case in point is Alex Mou, who‘s example I’ve followed recently. I now regularly help our Brazilian tableau friends with their questions.

However not everyone has the language skills or the confidence to bypass whatever language barrier is there to try and help. By allowing users from other countries the ability to post on the main forums and get automatic translations on the fly is a great addition to the community and something that will hopefully be used by a large part of the worldwide Tableau family.

Similarly, back in October I did a viz where I wanted to reach English and Portuguese speakers alike. To date this was probably my most challenging data visualization, not only because I got stuck and had to enlist help from one of the Zen Masters(more on that later) but because I’ve made some pretty specific considerations on the design and how I wanted the viz to flow.

So here’s the dashboard that greets you in Tableau Public. This was always meant to be clean and clutter free, soon after I published, I thought it would bet better to have the dashboard in one page only and have a parameter allowing language selection. In any case I think it works well as the greeting page. Go ahead and click it, it will open a new tab in Tableau Public where you can interact with it.

The technique used here is pretty simple; I created a worksheet for each of the flags and added them to a dashboard after which I added a dashboard action which upon clicking on the flag the user is taken to the respective dashboard.

You can read a more detailed explanation from Tableau’s Knowledge Base articles

Navigating to other dashboards

When it came to the infographic I wanted to try and mimic an infographic like the ones I sometimes see in newspapers. Yes it can be argued that a dashboard shouldn’t be scrollable; however in this case I decided to go against the rules. Additionally I split the dashboard in two themes. The upper side includes a quick overview in the shape of a bar chart with the countries where the rate of early school leaving is higher than the average are darker than the rest.That colour difference is found throughout the analysis.

This is achieved by creating a calculated field similar to the one below. “% Abandono Escolar” translates as “% rate of early school leaving”

After this I looked at how the rate of early school leaving is dropping across Europe and its percent difference since records were available. Things got really interesting from here on. Unfortunately as it’s always the case not all countries supply the same data from the same time periods. For instance Italy has data going back to 1992, whereas for Hungary Eurostat only supply data from 1997 onwards.

I looked at the differences between genders and found that more often than not men typically rank higher in early school leaving rates. To do this I did a bar in a bar which can be a very useful way of displaying the data.

Using superstore data just follow the instructions below:

  1. Bring Sub-category to rows

  2. Sales to columns

  3. Drag Profit on top of the axis until you see the symbol that represents a combination axis. (see below)

  4. From columns drag your Measure Names to colour

  5. Hold Ctrl to duplicate Measure Names and place it on size. You should end up with two instances of Measure Names in your marks card, one in colour and another one in size.

  6. Now for the “pièce de résistance”, go to analysis up top and turn off stack marks.

Your chart should look something like this.

These types of charts are handy when comparing target with year to date performance, or cost to budget, or when you are comparing a measure against another as it was the case in my analysis where I wanted to compare between genders.

Finally I wanted to rank men and women for the selected country against the other countries but only display the country selected and across its respective time period. There’s a great explanation from Emiliano Colosimo here:

A Jedi (Filter and Table Calc) Trick

Followed by Mark Lutton’s explanation on how you can make this work across various worksheets. This is done by using a parameter which will be a problem if you are trying to roll this out against fields that keep changing like customer names. In my case this wasn’t a problem so it worked perfectly.

Creating a Global LOOKUP(MIN(…),0) Filter with a Parameter for “Jedi” Filter and Table Calc Trick

and then I got stuck……

I came across a problem where I wanted my percent difference to be calculated from the year available for that particular country. Remember the country selector was a parameter and not a filter. It showed the country I wanted but by hiding the others not by filtering the actual data. As a result when selecting Hungary which only had data from 1997, my chart showed nulls across the range as it tried to calculate a percent difference from 1992, which was the year where data started for a number of countries. Tableau couldn’t find 1992 for Hungary therefore it assigned a null and the calculation was invalid.

I spent about a week trying to get my head around the problem until I spoke to Peter Gilks who kindly fixed the problem for me. The solution was simple (well once you know it), context filters.

So far I have had little experience with context filters and they have always come with a performance warning but in this case it worked perfectly. What Peter did was to filter the country before anything else. So if the parameter is equal to the country then apply the context filter to that worksheet only. Allowing me to use my parameter across the different worksheets adjust the starting point of the percentage difference calculation and use a rank chart where we hide the non-ranked countries.

Peter has since written a really good and detailed blog post on the subject of context filters which I really recommend.

Let’s put some things into context – a practical introduction to Context Filters in Tableau

I created the same version in Portuguese by duplicating fields and changing alias which was by far the easiest way I could think of and created a similar dashboard.

As I said in the beginning this was by far the most complex visualisation I did on tableau public and I learned a lot which was the main objective. Feel free to download the workbook to see the different calculations and how everything hangs together.

Thanks for reading, Merry Christmas and a Happy New Year

Feliz Natal!!!


#datavisualisation #opendata #tableau