Product Tour
forums
Use these forums to ask questions and discuss Tableau.
Displaying and calculating confidence intervals
In a trend plot, say, cancer rate by year, is it possible to display a confidence interval? (margin of error) or is it possible to display a high-low plot as is done with a stock price chart over time. Usually the confidence intervals are numbers that I calculate, although some canned ones (for means for example) would be useful too.
In Excel it is possible to put confidence intervals around the points of a trend line and confidence intervals on bar charts.
Can't see a way to do that in Tableau.
Since Tableau offers the capacity to use a variable in an aggregated form such as an average, it would be reasonable to be able to display the confidence interval for that mean as well as display confidence intervals.
Any ideas about to to display confidence intervals?
Subscribe to the comments on "Displaying and calculating confidence intervals"
Comments
Go to http://www.tableaucustomerconference.com/sessions/training/
and download the materials for the "Means, Modes, Percentiles & Outliers: Using Tableau to Understand Distributional Data" presentation.
You'll need version 4.0 to read the workbooks, but you can produce box plots and more. Confidence intervals have been in Tableau for a while.
There are some other good materials up there also.
I can find no reference to confidence intervals in the Tableau Help. Can you tell me where I should look?
Typically I calculate the confidence intervals I need, and would like to plot them with the point estimate. Excel has a stock high-low-close chart that is close but still not really what is needed to present statistical modeling or scientific data.
thanks -
You can show confidence intervals, quartiles, medians, min, max etc by creating reference lines (by right clicking on a continuous axis)
Reference lines are described in the help under
Home > Building Data Views > Inspecting Data > Reference Lines
The conference workbook mentioned above has great examples showing distributions. The most relevant example involves a few steps: such as turning off aggregate measures, lowering the transparency, turning on reference lines, and placing two copies of the discrete dimension on a shelf so that the reference lines are computed once for each discrete value of the dimension.
For this to make sense, you may have to download the example and play with it a bit to see how he built it. His example turned on reference lines for quartiles, min and max, but you could just as easily select a confidence interval instead. You can pick the confidence level, and let Tableau do the calculation.