0:00Hey, it's Tim here today, different kind of
0:02video, I'm actually uploading a correction.
0:04Yes. When I
0:05made a video about the error calculation in
0:072021.2, there was essentially a mistake
0:10that I made, and it
0:11was an assumption. Now what actually
0:13happened in this error calculation is I had
0:15a data set
0:15from a London data provider. And in that
0:18data set that actually pre calculated the
0:21area. And the
0:21reason I was looking for this data set is
0:23because I needed some way of validating
0:25tableau's error
0:26calculation against another data set. So
0:28the assumption I made was that the company
0:30that
0:30produced this data had done it correctly.
0:33Now it actually turned out that when I did
0:35this comparison
0:36in the video, you can see here at seven
0:38minutes 55, I'm actually comparing seven
0:40side by side,
0:41the two calculations, and you can see that
0:43I'm calling out that the tablet error
0:45calculation
0:46is way off, actually sent this feature into
0:48tableau. And they actually took some time
0:50out
0:50to look at the data set and to look at the
0:52error calculation. And it turned out using
0:54another tool
0:55that the error calculations in the data
0:57source I was using was actually incorrect.
1:00And therefore,
1:00I came to the conclusion that the error
1:02calculation was not as accurate as it
1:04should be
1:05over certain types of geographies. And in
1:07some cases, you can see here side by side,
1:09it maps pretty well. But there was just a
1:11few object IDs here that were generating
1:14areas that just weren't matching. And that
1:16wasn't the fault of tableau is actually the
1:18fault of the
1:18error calcs that were done in this data set
1:21. Now, it is really important lesson, which
1:23is to never
1:24make the assumption that the data source
1:26that you find from a trusted source is
1:27always correct.
1:28And number two, I jumped to a conclusion
1:30without first validating it before I
1:31released the video.
1:32Now in order to fix this, what I'm actually
1:34going to do is I'm going to play the video
1:35again,
1:35I'm not going to re record the video. But I
1:37will put a note up in the section where I'm
1:39calling out
1:40the mistake or the issue with tableau's
1:41data source to actually validate validate
1:43that the
1:44issue here is not to do a tableau. And I
1:46sort of don't want to re record the video
1:48and then delete
1:49the old one because I think it's good to
1:51sort of fess up to your mistakes. And
1:53actually realize that
1:53we're all still learning. And at the end of
1:55the day, tableau did the due diligence and
1:57to make
1:58sure that the calculation was doing the
2:00best job it can. Now, if you're into some
2:03sort of detail
2:04about this, it's actually quite hard to do
2:06error calculations on something like the
2:08curvature of
2:09the earth, depending on the size of
2:11geography, you can get slight variances
2:13depending on how you do
2:14the calculation. So I'm not the expert on
2:16this, I'm going to maybe link to some
2:17articles in this
2:18description for this video. So you can go
2:20off and find out the complexities of doing
2:22that there's
2:23actually a tableau paper on hexagons and
2:25hex mapping and sort of showing the
2:27disproportionate
2:29behaviors that you can get and how you have
2:31to sort of correct for that over large
2:32areas. So I
2:33actually recommend you read that that's a
2:35proper in depth paper by people who know
2:36what they're
2:37talking about. And that can give you more
2:39context into this. But that's pretty much
2:41it for me,
2:41I just wanted to add this correction in
2:43front of the video, but upload the video as
2:45is rather
2:45than trying to sort of hide the fact that I
2:47got it wrong. So there we go, we all make
2:49mistakes. And
2:50hopefully you can learn from mine. Thanks
2:52for watching, and I'll catch you in the
2:53next video.
2:54Hey, Tim here in today's video, I'm taking
2:56you through a new feature in 2021.2. And
2:59that is the
2:59new area function, a new spatial function
3:02that allows you to calculate the area of a
3:04polygon.
3:04But before I get into this, be very careful
3:06with this. If you use this over large ge
3:09ographies,
3:10it's actually only giving you an estimate.
3:12I'll get into why that is in a second. Let
3:14's get stuck
3:15into this. So the first thing I'm going to
3:16do is I've actually gone to this website,
3:18when you're
3:18working with spatial data, it's always good
3:20to have a good sample data. And I've gone
3:22to this
3:22London data store to grab flood risk zones
3:25as calculated by the Environment Agency.
3:28Essentially,
3:28they figure out where there's a flood most
3:30likely to happen in London. And they've
3:32created a shape
3:33file or a bunch of polygons for us to use
3:35if we want to visualize that it's part of
3:37an open data
3:38initiative. Now, this was created three
3:40years ago, so it's nothing new. But I'm
3:41going to connect to
3:42this directly in Tableau, I've actually
3:44gone ahead and downloaded it to my desktop,
3:46so it's already
3:46there. So I'll just go back into Tableau
3:48here, and we'll just go ahead and use the
3:50spatial file. So
3:51if you've never used a spatial file before,
3:53it's actually this option here, it's
3:54essentially a type
3:56of file that has information about
3:57geography, that's the simplest way to sort
4:00of think of it.
4:00And so if I click on this, you'll see that
4:02I actually have the ability to open the
4:04file and
4:05shape files don't just come on their own,
4:07they tend to come with other supporting
4:08files. But the
4:09one file that we're interested in is this
4:11one, which ends in dot s HP. So if we click
4:14that and
4:14click open, you should see that Tableau
4:16handles this absolutely fine. It's been a
4:18native capability
4:19of Tableau to support these out of the box.
4:21So it's just right into this Tableau. And
4:23the great
4:24thing is that it works with everything
4:26including the data models. If you want to
4:27pair this up with
4:28other days that you have and find out where
4:30the properties are in certain London flood
4:32zones,
4:32you could it's an absolutely very simple
4:34thing to do. Maybe that's for another video
4:36. But
4:37nevertheless, we've connected to this data.
4:39Now I'm just going to go right into sheet
4:41one. And
4:41you can see that here on the left hand side
4:43, we've got all our fields. Now in Tableau,
4:46it's super
4:46easy to create a map, you just double click
4:48any geographical field, and it will do it
4:50for you.
4:51Where I know it's geographical is because
4:53there's this world icon to the left. So let
4:55's just double
4:56click that and see what happens. It's a
4:58geometry field and boom, we already have a
5:00map. Now, I know
5:01this is London, but it has some really poor
5:03context on here. There's nothing on the map
5:05that suggests
5:06this is London. So let's customise this map
5:08before we carry on. Let's go to map layers
5:10at the top.
5:11Let's choose a more interesting map, let's
5:13choose streets. And this should start to
5:16give us a little
5:16bit more sort of a vibrant map. I'll sort
5:18of dial down the the wash in the background
5:21so that the
5:22visualisation is actually what we see
5:24prominently. But it's nice because I have a
5:26literal label that
5:27says London, have all the London boroughs
5:29and towns. So if you're familiar with
5:31London, this
5:31should start to look very familiar for you.
5:33But nevertheless, we're here to look at
5:35this new
5:35function. So I'm going to close this map
5:38layers option. And when I look at this data
5:40, if I just
5:40open this little data preview for that, I
5:43just went over here, that clicking that
5:45little table
5:46icon opens up this data preview that you
5:48can see. And you can see that there's a
5:49bunch of different
5:51shapes, and each one has an object ID. So
5:53what I'm going to do is I'm actually going
5:55to use that
5:55object ID. Now, the other thing to bear in
5:58mind is that I've actually already got the
6:00area in here.
6:01And the reason I want this in here is
6:02because I want to show you how accurate
6:04this area calculation
6:05actually is. In essence, the bigger the
6:08proportion of area, the less accurate this
6:10is going to be.
6:11And we'll kind of do a quick calculation
6:12just to see how much it sort of varies. So
6:15let's,
6:15let's go ahead and look at that. So let's
6:17go ahead and let's just close this. And now
6:20what I'm going
6:21to do is grab objects ID and just put it on
6:23colour so that each and every one of these
6:25polygons gets
6:26its own sort of colour. Now, the reason
6:28this isn't quite showing what I wanted to
6:30show is because I
6:31need to bring in more context. So what I'm
6:33going to do is I'm going to actually drag
6:35the object ID
6:36up into this table section. What's
6:38happening here is that the object ID is
6:40being treated as a number.
6:42So it's actually aggregating it. I don't
6:44want that. I want it to treat it as like a
6:45text field. So
6:47what I'll do is I'll cheat on it a little
6:48bit. I'll actually drag object ID up here.
6:51You'll see
6:51that it changes two dimensions when I do
6:54that. And now it's blue. And now what I can
6:56do is put
6:57that on colour. Now each and every one of
6:59those shapes has its own unique colour. And
7:02if I hover
7:02over that, you can see that it sort of
7:04activates and we're now able to click on
7:05them and interact
7:06with them. So that's working really, really
7:08well. Now, the other thing I'm going to do
7:10is I'm also
7:11going to just look at this table again, and
7:13just see what's also useful to bring
7:14probability band.
7:15That's probably telling you how much of a
7:17flood risk zone this is. So what I'll do is
7:19I'll put it
7:20on a tool set because I want this in the
7:22tool tip when I hover over something. So
7:24you can see that
7:25some of these are very low, some of these
7:27are very high. And it essentially changes
7:29depending on the
7:30part of the map that we click on. But it's
7:33sort of working. Anyway, we've got enough
7:36information
7:36on our visualisations. Let's go ahead and
7:38actually use this function. So it's a super
7:41simple function.
7:41It's actually one of these functions in
7:43Tableau all together. So if I just type in
7:46area, you see
7:47that you get a little hint here of what you
7:49should use. So you should use a polygon
7:50dimension in here
7:52and the units that you want to measure the
7:53area in. So in this case, my polygon
7:55dimension is this
7:56geometry field. So I can actually just drag
7:59it in and drop it right there in Tableau,
8:01then hit a
8:01comma. And I'd like this done in metres. So
8:04I'm just going to do two speech marks
8:05single or double
8:06works perfectly fine. Enter an M and then
8:09now that's our calculation done for metres.
8:11You can
8:12also do kilometres and miles, you just have
8:14to find the appropriate documentation for
8:16that. In
8:17fact, let me bring that on screen now
8:18actually, you know, when I do these videos,
8:20I literally
8:21have the documentation to the left of me.
8:23So you can see here, the different
8:25dimensions you can use
8:27so metres, kilometres, miles and feet,
8:29those are all the different syntaxes you
8:31can use. So you
8:32could actually probably use a parameter to
8:34switch between those. So you can calculate
8:36all of them
8:38and then just put a parameter in here to
8:39change metres to feet and then it will
8:40dynamically
8:41recalculate it rather than using a
8:43parameter to change the different
8:44calculations. But nevertheless,
8:46this is really cool. And we've got the area
8:49in metres squared, I just put empty as a
8:52shorthand,
8:53and we'll hit apply. And now what I want to
8:55do is I want to grab the area in metres
8:56squared and the
8:57shape area and put them both on the tooltip
8:59so we can actually see what those two are
9:01like. Okay,
9:02now that's working, all the calculation is
9:04done. And now when I hover over this, you
9:07can actually
9:07see the two values, you can see here area
9:09in M2 and the shape area. Now there are
9:12subtle
9:12differences, okay, and we're talking
9:14roughly 1000 metres squared here or there,
9:17depending on how big
9:18the geography is. If I go down here to a
9:21slightly smaller one, you'll notice that
9:23the differences
9:24become larger, the bigger the geography,
9:26but if we go to a really small one, let's
9:28go to one of
9:28these ones here, they're very, very small
9:30differences. So essentially, the problem
9:33with
9:33this function is that the shape of the
9:36world is not flat. If you're a flatter,
9:39this is not the
9:40video for you. But essentially, the world
9:42is a curved platform. And the more curved
9:44something is,
9:45the harder it is to actually calculate the
9:47area because you've got to sort of flatten
9:49it out. And
9:49by the time you do that, it doesn't
9:51actually make a perfectly flat shape. So
9:53what this area
9:54calculation doing is sort of trying to
9:56compensate for that, but it's not really
9:58doing it sort of
9:58fully, you need a little bit more of a
10:00complicated sort of calculation in order to
10:02do that. Whereas
10:03the error calculation in this spatial file
10:05has actually been calculated by the
10:07environment
10:08agency using fairly accurate mechanisms,
10:10because they have to know the size of these
10:13things,
10:13things like insurance policies are based on
10:15these ratings. So they have to be pretty
10:17accurate as to
10:18how they work. So you can see there's a big
10:20difference. And what I'm going to do is I'm
10:22actually just going to build a very, very
10:24simple visualization, I'm just going to
10:26bring in object ID
10:27at all of them. And what I'm going to do, I
10:29'm going to bring in the area in meter
10:32squared,
10:33and I'm going to bring in the shape area in
10:34the table next to it. So you can just sort
10:37of see
10:37some of these side by side. Now, some of
10:39them are very similar, some of them are
10:42very different.
10:43For example, this one is sort of
10:44interesting, because, you know, this one's
10:47just so way off. So
10:48why is Tableau getting that so different,
10:50whereas some of these, they're almost so
10:52bang on. So it
10:53could be to do with the shape and the form
10:55of the particular geography could be a long
10:58versus thin
10:59versus, you know, you don't know all these
11:01sort of variables. So in generally speaking
11:03, you can see
11:04that this is an estimate. But it's just
11:06worth being careful and knowing that this
11:08is an estimate,
11:08because you see that this particular object
11:10, it says it's 12. But the actual area is
11:13calculated
11:13by the environment agencies 224. So that's
11:16just something to be aware of. Now, what
11:18you could do
11:19is you could actually sort of create a
11:21table. So if I was to sort this from the
11:24largest shape area
11:26to the smallest, you can see that some of
11:28these are just way off. So it's definitely
11:31one to be
11:31sort of wary of and to be really careful
11:34how you're sort of relying on this function
11:36,
11:37because it can actually throw some of your
11:39things off if you rely on it too much. What
11:41are the best
11:41use cases? Listen, if you're measuring
11:43things like playgrounds, parks, this is
11:45perfectly fine,
11:46you know, nice, simple geographies, polyg
11:49ons, shapes, simple geographies, if you're
11:51measuring
11:51the area of let's say, a river that travels
11:53over a long distance, well, that's going to
11:56be a little
11:56bit more tricky, because of course, over a
11:59longer distance, there's more curvature.
12:01And the shape
12:02area or area calculation is only going to
12:04be able to calculate it to sort of one
12:05extreme. So you
12:06will get less accuracy for polygons that
12:09span really large distances and large ge
12:12ographies.
12:13But I'm not a scientist here, I'm just sort
12:15of going off what what I see and what I
12:17know I've
12:17asked the devs about. But it's definitely
12:19one to just watch out for. And some of
12:21these I think I
12:22think there's probably just an issue with
12:23the polygons, you can see that that says 20
12:26. And then
12:26it's just completely way off. So it's not,
12:28it's not sort of an exact science. And I
12:31need to get
12:31more into this to really understand, you
12:33know, why are some of these really, really
12:35close and why
12:36some of them really, really far off. The
12:38figure on the right is the accurate one
12:40from the Environment
12:40Agency, I just wanted to get that there
12:42before I show you a calculation. And I tell
12:44you that it's
12:45correct as fact and then he has a slight
12:47margin. And actually, some of these have
12:49huge margins. So
12:50I'd be interested to see what you think of
12:51that. If you're a geographer, or you
12:53actually use spatial
12:54functions all the time, I'd love to know.
12:57And if you know anything about why this is,
12:59why is this
13:00calculation in Tableau getting it so
13:02different to the values that have been
13:03calculated by the
13:04Environment Agency, what I will do is I'll
13:06put a link to this data set and maybe you
13:08can tell me
13:09if you've got some insights to why that is.
13:11But I'd love to get to know more. Alright.
13:13Anyway,
13:14that's that's it. We have arrow functions
13:17in Tableau. Now you can use these over
13:19simple
13:19geographies and they're going to be really,
13:20really useful. Thanks for watching. If you
13:22've enjoyed
13:23this video, be sure to check out some of
13:24the other videos in 2021.2 on this channel,
13:27as well as other
13:28versions of Tableau from way past all the
13:30way to functions in Tableau. I cover
13:31everything so get
13:32in touch. Let me know what you'd like to
13:34see and if not, hopefully I'll catch you in
13:36the next one.
Tableau | Data visualisation | Analytics