I am currently exploring Tableau and I wonder if it can do the following.
Let's say I have a table with two columns, A and B. Let's say A can attain discrete values from a small set (maybe 10 different values), while B has continuous values. I would like to plot a box plot where on the x-axis are the 10 possible values of A, and on the y-axis the box plots. Each box plot shall contain values from B for a given value of A.
In case I have not made myself clear enough, here is an example. Consider data containing average temperature on each day within one year, labeled by week days. That is, we have 365 pairs (A, B), where A is a value between 1 to 7 (day of week), and B is the temperature. I would like to have 7 box plots, each containing data for the particular day week. (To prove that on Mondays there is always bad weather :))
Changing the problem slightly to assume the columns are named Date and Temp, and that Date is a real date with a month, day and year. On the data pane (left margin), make sure the Date field is a discrete (blue) dimension with datatype date (not string).
Drag Temp to rows, make sure it is a continuous measure. I would choose the aggregation function AVG instead of Sum for temperatures.
Change the mark type from automatic to circle, and then drag Date to the detail shelf. Right click on the Date pill you just dropped and choose Exact Date.
Change to the Analysis table and drag in box plot from the Summary section. Format to taste.
To get one column per weekday, drag Date onto the columns shelf. Right click on the Date field on the columns shelf and choose Weekday from the context menu (submenu)
Related
Spotfire question: I have a data table with monthly data that is visualized in a bar chart. I however also want to visualize the information in quarters, by showing the latest month of the quarter in the format '20Q1, 20Q2, etc.'. (So I don't want to use standard 'date format'.)
My idea was to create an additional column that is filled for March, June, Sep, Dec and empty for the other months. Then with a document property, the user can select to either view the data in months or in quarters (i.e. the last month of the quarter).
So far so good, my data now looks like this:
Month
Value
YearQuarter
Jan-20
100
Feb-20
200
Mar-20
400
20Q1
Apr-20
125
May-20
101
Jun-20
300
20Q2
The problem now is that when I visualize the data with YearQuarter on the x-axis, it also shows all the (empty) values in a bucket. See below. How to solve this? Note that the x-axis has a custom expression "<$esc(${Granularity})>", where Granularity is a document property to determine what column to pick.
Did you try limiting your data with whatever expression you have put in x axis.
Thanks
I've created a parameter A and calculated field B to choose filter (between C and D) based on A.
How can I set filter value, when make a choice in B?
Something like:
IF B==C then C = 'enabled'
IF B==D then D = 12
Field C and D have different dimensions, so, making choice in B and then switching to another value
I got an empty dashboard before setting any value in the filter.
When I choose Year to date filter and switch to ENABLE I see my plots, but when changing to By month option nothing selected, and to fix any visualization I need to choose values again. When I switch Year to date this story returns.
When I choose Year to date filter and switch to ENABLE I see my plots
but when changing to By month option nothing selected, and to fix any visualization I need to choose values again
This is the way how I calculate the chosen filter
i'm totally new to Tableau but that is what I could potentially use at the workplace, so asking this question to decide if its worth it.
I've a monthly values dataset and I'd like the tool (Tableau) to generate a report to point out anomalies - the ones i have in mind right now are:
1) same data value for x months in a row
2) data value is 0
3) 5 parameters been reported last 3 months and all of sudden only 4 are reported
Is this possible in Tableau?
You can certainly do this. The below is a general example for doing anomaly detection (one standard deviation outside of mean) but you can modify the calc to fit your criteria. If you place your date field in the column position and your metric, lets say Sales, in the row position you can then create this calculated field to label the anomalies.
IF SUM([Sales]) < (WINDOW_AVG(SUM([Sales])) – WINDOW_STDEV(SUM([Sales]))) THEN “Bad
Anomaly”
ELSEIF SUM([Sales]) > (WINDOW_AVG(SUM([Sales])) + WINDOW_STDEV(SUM([Sales]))) THEN
“Good Anomaly”
ELSE “Expected”
END
You can then place the new field on your color mark to highlight each of the results.
I'm working with a dataset that describes contract durations with a list of companies. Obviously all contracts have a start date and a (minimum) duration. Some contracts however, are subject to possible prolongation. Here's a data example:
Company contractStart minDuration firstProlong secProlong
Company x 27-11-2011 27-11-2014 27-11-2015 27-11-2016
Company y 29-11-2014 29-11-2015
Company z 01-04-2011 01-12-2011 01-12-2015 01-12-2017
I want to visualize them in a Gantt'ish type diagram that has duration along the x-axis, and shows the respective contract lengths with the possible prolongations as a sort of stacked bar chart. Along with that I want something to mark the current date. I've managed to do this in Excel. See picture:
.
I would however really like to be able to the same thing in Tableau, and I really can't figure out how to. I've been working with two different approaches so far:
1st approach
Here I drag the Company to the Rows shelf, and the contractStart date to the Columns shelf. I then create a calculated field that calculates (with DATEDIFF) the difference in days between the contract start and the minimum duration. I can drag that to the Size shelf, set the mark type to Gantt bar, and get something like this:
I can then go on to calculate the date difference between the minimum duration and the 1st prolongation and the same for the 1st and 2nd duration. The problem here is that I can't add these calculated fields to the Size shelf as it can only hold one at a time.
2nd approach
Here i reorganize my data like this:
Company Type startDate endDate
Company x minDuration 27-11-2011 27-11-2014
Company x firstProlong 27-11-2014 27-11-2015
Company x secProlong 27-11-2015 27-11-2016
Company y minDuration 29-11-2014 29-11-2015
Company y firstProlong
Company y secProlong
Company z minDuration 01-04-2011 01-12-2011
Company z firstProlong 01-12-2011 01-12-2015
Company z secProlong 01-12-2015 01-12-2015
I then create a calculated field to find the difference in days between startDate and endDate. I drag that to the Columns shelf and drag Company to the Rows shelf. I then drag Type to the Color shelf, set the mark type to Bar, and get something like this:
The problem here is that I need dates along the x-axis not integers. Also I need to be able to sort the chart like my initial Excel example. I.e. I want to sort on the longest possible duration of a contract (ascending). Also having to reorganize all my data seems a bit tidious.
Here's the spreadsheet and the tableau file I've been working with. Am I overlooking something really simple?
Your approach 2 is close. instead of sum(datediff) on the columns, place StartDate as continuous. Then place sum(datediff) on size. You can now sort Company ascending on sum(datediff). Sometimes, reorganizing data is necessary.
This relates to my How do I achieve a pivot report in Crystal Reports for Visual Studio? The data is the same, but they want a line chart as well.
So, my data is {datetime}, {car-id}. The X -axis of the chart must show the hour of {datetime}, the Y axis the number of cars that entered the park at that hour, and I must have seven data series, lines, one for each day of the week. So, e.g. the x/y point on the green line shows that on Wednesday, at hour x, y cars entered the car park.
Common sense tells me that I can kludge this by transforming the data source so that each day has its own column in a table, a table like this:
DateTime
WeekDay
CarCount
Yet the Crystal line chart doesn't seem to support more than one column, so there must be a different and better way of doing this. The weekday is part of {datetime} after all.
How can I achieve this chart? I am a rank amateur at charting, and Crystal's idiosyncrasies really aggravate my lack of skills and experience in this area.
It's not obvious how to get multiple series to show up properly.
First, create a formula that will extract the day of week from the datetime: dayofweek({datetime}) and then create another that will extract just the hour: hour({datetime}).
From there, insert a chart from scratch and make the following settings in the Chart Expert
Select "Line" under the Type tab
Under the Data tab, select "On change of" and select your {#HourOfDay} formula first. This will create each hour as a point on the X-axis.
In that same spot, also insert the {#DayofWeek} formula. This will create a new line
for each day of the week. This is, in my opinion, the non-obvious part.
Finally, to pick your Y-axis values,
select {car-id} where it says "Show value(s)". It should default to
the count summary of that field.
(Note: in the screenshot below, you would just replace {Customer.Customer_Name} with {car-id}
I had this issue for 5 years and finally solved it. I have a XY chart with multiple series.
Everything needs to be in a formula.
The graph should be in the report header section of a subreport. The sub report generates the X Y data like this:
Series X Y
Ser1 2020 1
Ser1 2021 2
Ser2 2020 0
Ser2 2021 1
For me I have as series "Energy" ie 6, 9, 12 etc.
Then I have a date/time and a measurement.
On change of (Energy formula)
Show Values
#1 DateTime (convert this to int ie 2000.9 for 9/2000) - dont summerize
#2 Measurement (number) - dont' summerize
Then you can do the color highlighter to capture your series (6 is red etc).
The legend won't work - can't figure out that yet.