Applying Aggregated Measures to a Box Plot in SAS VA

[SaiKrishna3]

I need to use aggregated measures in a box plot to make the data sensible - and it seems odd that I can't use that in visualization?

Can anyone explain why it is not possible to generate a box plot using aggregated measures? If no, Is there any alternative way that i can convert aggregated measures to measures so that i can use it in my box plot.

Regards

Sai Krishna

[MarkusWeick]

Hi @SaiKrishna3,

could you please provide an example of what you want to achieve.

(Box plots are a way to show the distribution. After aggregation most of the distribution features are gone.)

Thanks

Markus

[SaiKrishna3]

Hi @MarkusWeick ,

Appreciate your quick response.

In my example: I have created a new calculated data item in SAS VA which is assigned as Aggregated Measure.

Now, I am unable to add this aggregated measure to a newly created box plot, since there is no such option to create one. There is just "Measures' option.

Is there any way that i can use aggregated measures to a box plot? (or) is there any alternate way to change the aggregated measures to a measure(there is no option to convert agg measure to measure)  to use in a box plot.

Regards

Sai Krishna

[MarkusWeick]

Hi @SaiKrishna3,

in my experience the answer to your two questions (Is there any way that i can use aggregated measures to a box plot? (or) is there any alternate way to change the aggregated measures to a measure(there is no option to convert agg measure to measure)  to use in a box plot) is "no". I think you need to do the calculation in the preprocessing of th data before you load it into VA.

(The only "grouping" in a box plot is by the variables of the category. And if you aggregate over this "group" there would be only a single value, for which a box plot wouldn't make sense.)

Sorry for this unsatisfactory answer. But maybe @Madelyn_SAS or @Stu_SAS have a better one.

Best

Markus

[Stu_SAS]

Hey @SaiKrishna3! It looks like you're trying to apply an error percentage, but you'd like to do it by group. The best way to do this is to calculate your error percentage as a standard, unaggregated measure:

('Actual'n - 'Predict'n) / 'Actual'n

You can then create a box plot for each by-group with the Category or Lattice data roles. Box plots do not support aggregated measures. Aggregating data and applying it to a box plot will produce a single point, and aggregating a box plot to a specific by-group will create a set of points for each group but they will not necessarily be helpful for interpreting your percent error as a single box plot. In that case, you're better off using a bar chart grouped by a category to get the average error of each group.

If you want to focus only on a specific by-group, it is best to add a filter / control that lets you select the specific group of interest.

My personal recommendation is to use a histogram for interpreting error for normality assumptions and reviewing extremes.