Forum Discussion

Anonymous's avatar
Anonymous
Not applicable
7 years ago
Solved

How to plot box and whisker?

Hi 

 

I am trying to plot the min, max, median, 1st and 2nd quartile in Power BI. I thought a box plot would be the best, so I imported the box and whisker chart by Jan Pieter and the and the violin plot. I am confused about how to use these custom plots. Mainly at the Value setting, as usual for numerical fields I can choose to summarize it as max, min, median, mean etc, but I don't understand why this is the case. I thought the whole purpose of the chart is to generate those at once in the chart. I created a table with measures where I calculated the max, min, median and so on for reference. But the numbers don't match. Please see the snapshots where I used either Sum or Average to plot the charts. The numbers won' match my calculations. Why? How can I fix it?

 

 Box and Whisker:

'Sum' applied.

Average applied:

 

 

Violin Chart:

Sum applied:

Average applied:

 

  • dax's avatar
    dax
    7 years ago

    Hi wsspglobal

    According to your description, it seems that you want to understand the logic of above chart, right?

    For whisker, you could find that it group on category, then calculate  the max of sampling

    For plot, you could find that it also  group on “Category”, and calculate max, min based on “sampling”

    So you could modify field in “Category ” and “sample”  based on your requirement.

     

    Best Regards,

    Zoe Zhi

     

    If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

  • Hi Anonymous - I just came across this post while browsing - I'm the author of the violin plot visual and may be able to help clarify its usage, if you're still experiencing some challenges.

    This is documented here, but I'll walk through using your data.

    You are correct that the visual should create all statistical measures, based on the data you feed into it. In its simplest terms, you need to be able to provide data at its lowest possible grain for this to be worked out correctly. If any of this data is pre-aggregated then it will not work.

    From your example, you are plotting Session as a Category and Amount as the Measure Data. If we only pass these in the the visual, then Power BI will aggregate Amount by Session - as you've already indicated in your screenshot - but to illustrate how this looks to the visual:

    You will notice that the Amount is aggregated by the Session. In order for the visual to know more about the distribution of your data, we need to give it every row we can and this is where the Sampling field comes in.

    You need to create a unique value for this field so that the visual does not aggregate the measure at a higher level. We'll look at your data and try and find a suitable candidate field.

    An easy way to check this is by adding a measure that provides a count for the one you want to check. Let's have a look at Subset:

    Note that there is a 2 in each column. This indicates that Subset is too high-level to provide all values of Amount to the visual, and we can see the resulting violin plot:

    The tooltip says there's 5 samples, because there are 5 distinct values for Subset, and we actually need 10 because that;s what's in your raw data.

    An ideal way to do this is to use something like a primary key in your table to do the sampling on. The puts the data at its lowest level of grain. You can do this in Power Query, e.g.:

    1. Add Column
    2. Index Column
    3. From 1

    This creates a unique value for every row in your table, e.g.:

    Close & Apply the query and this will get added to your table. Let's repeat the above 'Count Test' using Index:

    We can now see that every row has a 1 in it, which means that for every row, the Amount is going to be as per the table and won't pre-aggregate, e.g.:

    Now, we should add Index to the violin plot so replicate this behaviour, e.g.:

    If we check the tooltip for A, we get the stats calculated, e.g.:

    This technique will also work for the Box & Whisker Chart, e.g.:

    Hopefully this helps clarify the usage and behaviour of both of these visuals, but feel free to ask anything else and I'll do my best to answer.

    Good luck!

    Daniel

5 Replies

  • Anonymous's avatar
    Anonymous
    Not applicable

    Here's my data:

     

    SessionSubsetYearAmount
    Aa201181.82676
    Ab201138.07561
    Ac201157.11986
    Ad201134.58256
    Ae201158.59327
    Ba201117.83215
    Bb201176.71251
    Bc201123.25766
    Bd20115.124324
    Be201155.95021
    Ca201112.54243
    Cb201127.4061
    Cc201172.05263
    Cd201165.40645
    Ce201199.42319
    Da201188.69064
    Db201197.62212
    Dc20114.266029
    Dd201143.47741
    De201167.51566
    Ea201126.86167
    Eb20114.488669
    Ec201125.73854
    Ed201176.41257
    Ee201120.14493
    Fa201140.40302
    Fb201185.39068
    Fc201140.93762
    Fd201142.4243
    Fe201132.9475
    Aa201266.79593
    Ab201293.1857
    Ac201258.68907
    Ad201242.2477
    Ae201277.47958
    Ba201275.50514
    Bb201270.82896
    Bc201256.12918
    Bd201210.55485
    Be201276.93667
    Ca201218.18667
    Cb201263.1883
    Cc201293.8165
    Cd201281.05533
    Ce201275.86873
    Da201274.05272
    Db201242.40564
    Dc201230.16113
    Dd201257.68567
    De201240.86473
    Ea201242.71101
    Eb201265.83886
    Ec20128.514872
    Ed201219.7125
    Ee201224.29333
    Fa201271.20736
    Fb201280.1227
    Fc201286.54642
    Fd20129.776906
    Fe201277.82144
    • dax's avatar
      dax
      Community Support

      Hi wsspglobal

      According to your description, it seems that you want to understand the logic of above chart, right?

      For whisker, you could find that it group on category, then calculate  the max of sampling

      For plot, you could find that it also  group on “Category”, and calculate max, min based on “sampling”

      So you could modify field in “Category ” and “sample”  based on your requirement.

       

      Best Regards,

      Zoe Zhi

       

      If this post helps, then please consider Accept it as the solution to help the other members find it more quickly.

      • dm-p's avatar
        dm-p
        Super User

        Hi Anonymous - I just came across this post while browsing - I'm the author of the violin plot visual and may be able to help clarify its usage, if you're still experiencing some challenges.

        This is documented here, but I'll walk through using your data.

        You are correct that the visual should create all statistical measures, based on the data you feed into it. In its simplest terms, you need to be able to provide data at its lowest possible grain for this to be worked out correctly. If any of this data is pre-aggregated then it will not work.

        From your example, you are plotting Session as a Category and Amount as the Measure Data. If we only pass these in the the visual, then Power BI will aggregate Amount by Session - as you've already indicated in your screenshot - but to illustrate how this looks to the visual:

        You will notice that the Amount is aggregated by the Session. In order for the visual to know more about the distribution of your data, we need to give it every row we can and this is where the Sampling field comes in.

        You need to create a unique value for this field so that the visual does not aggregate the measure at a higher level. We'll look at your data and try and find a suitable candidate field.

        An easy way to check this is by adding a measure that provides a count for the one you want to check. Let's have a look at Subset:

        Note that there is a 2 in each column. This indicates that Subset is too high-level to provide all values of Amount to the visual, and we can see the resulting violin plot:

        The tooltip says there's 5 samples, because there are 5 distinct values for Subset, and we actually need 10 because that;s what's in your raw data.

        An ideal way to do this is to use something like a primary key in your table to do the sampling on. The puts the data at its lowest level of grain. You can do this in Power Query, e.g.:

        1. Add Column
        2. Index Column
        3. From 1

        This creates a unique value for every row in your table, e.g.:

        Close & Apply the query and this will get added to your table. Let's repeat the above 'Count Test' using Index:

        We can now see that every row has a 1 in it, which means that for every row, the Amount is going to be as per the table and won't pre-aggregate, e.g.:

        Now, we should add Index to the violin plot so replicate this behaviour, e.g.:

        If we check the tooltip for A, we get the stats calculated, e.g.:

        This technique will also work for the Box & Whisker Chart, e.g.:

        Hopefully this helps clarify the usage and behaviour of both of these visuals, but feel free to ask anything else and I'll do my best to answer.

        Good luck!

        Daniel

  • Anonymous's avatar
    Anonymous
    Not applicable

    Hi dm-p,

     

    I have stumbled accross this post and i found it extremely helpful for using the box and whisker plot shown above.

     

    However, I am having a problem where the mean and median calculated with the built in agregation functions are not the same as the mean and median in the box and whisker chart.

     Here we can see that the aggregations I have applied to "EPGcalculation" do not equal the same as those in the box and whisker chart.

    Can you tell why this is happening?

     

    Here is the link to mybpix file. all the data is from "bireportdata PhenRQCPlusData" and the visuals are on "Page 1".

    https://techiongrouplimited-my.sharepoint.com/:u:/g/personal/ben_techiongroup_co_nz/EQw8qurSCo9Oo0SRNoKEcYIBrI2Vz6w8AnEQFm8-jT72bQ?e=2V1Rpg

    Thanks in advance