4 Replies Latest reply on Apr 19, 2018 4:33 PM by Luciana Junqueira

# How to show lack of statistical base on survey data

Hello,

On this survey dataset, I have 5 cultural activity displayed by gender female and male.

Here whichever count of response I get is valid but If a second filter is applied to the gender results and the count number is below 30 respondents than the graph has to show that for the specific result it has lost its statistical basis.

For the caption I wrote a simple formula :

IF COUNT([GOING group]) < 30  THEN "▼" ELSE " " END

The problem is that the ▼ icon will appear for any count below 30 but it should only appear for the second filter result; results by X filter divided by Y filter.

Over the past year, I've worked with this survey data from 12 cities in Brazil.

I have watched many of the introductory "how-to" videos and search for answers here in the community but not many calculations can be applied to survey data.

I greatly appreciate any help or suggestions,   I look forward to further learning what is possible in Tableau

Thanks,

Luciana

• ###### 1. Re: How to show lack of statistical base on survey data

Hello Luciana

Is the indicator to appear regardless of which filter is selected or only based on the male/female filter?  For example - in the example workbook included, the schooling was set to "High School" for Schooling and i wanted to verify because looking at this, I would think the only impacted column would be Concertos in the sample data.

Patrick

• ###### 2. Re: How to show lack of statistical base on survey data

Hello Patrick

Thank you for your interest and help.

The indicators should appear regardless of which filter is selected. We have 10 filters and we let the users decided which ones to use. If they select many filters the survey shows data that has lack of statistical base and then we need to make this very clear.

Josh Davis helped us with a calculation using WINDOW_SUM to add together sums of values across parts of the view, so if they are <30 it will show the symbol.

I think this is a very important part of surveys so if Josh doesn't post it here I will place the workbook for further reference.

Luciana

• ###### 3. Re: How to show lack of statistical base on survey data

Luciana, I'm glad I could help!

I will let you post a workbook of your choosing since we keep the data you provide to us private

Although many calculation problems can be solved with either Table Calculations or Level of Detail Expressions, this question is a great example of something only Table Calculation functions can achieve: dynamically evaluating at a slightly broader granularity than the view.

If we needed to calculate across a specific dimension, we could use

{EXCLUDE [dimension to ignore]: SUM([measure])}

or

{FIXED [other], [dimensions], [in], [view]: SUM([measure])}

However, when the dimension to evaluate across changes, we need a Table Calculation.

Compute using Pane tells Tableau to evaluate the Table Calculation along the groupings of marks in the view, in the direction specified.

Pane (Across) groups all the marks within the second-to-last dimension on the Columns shelf, ignoring the last dimension on Columns.

Our article on Addressing & Partitioning contains further details on what is meant by a pane in the context of Table Calculations.

While we're on the topic of survey data, it's also worth taking a look at this whitepaper on Visualizing Survey Data

Thanks for sharing! Have a wonderful evening!

1 of 1 people found this helpful
• ###### 4. Re: How to show lack of statistical base on survey data

Attach workbook with the solution.

Thanks, Josh Davis