I think it would be phenomenal to be able to add a “likely” value to ranges.
For example:
we can already set an annual growth rate to 20% to 50%
This causes at least line charts to display a nice feathered range below and above the mid-point defined for the individual graph. (side-note: bar charts only seem to display the upper part of the range, between mid-point and the upper boundary of the selected uncertainty interval, however it “swallows” the lower end. Here it would also be great to have the lower boundary indicated e.g. by a small dotted line in the solid piece of the bar)
Now how great would it be to set a range such as 20% to 50% likely 45% and thus have the mid-point be defined straight in the formula definition.
You’d be able to transport much more certainty about your estimation ranges, plus the mid-point could change over time rather than being a constant “whatever in between”.
Right now we produce symmetrictriangular distributions, your proposal is to support asymmetric triangular distributions as well, which makes a tonne of sense.
We’ll have a think about the best way to add this. In the meantime, you can actually do this to accomplish roughly what you’re looking for:
(20% to 45%) to 50%
If you create a Distribution chart of a variable with this formula, you can see that the ‘mid-point’ ends up being 45%.
I do see the shift in the distribution curve, however it doesn’t seem to have an effect on the result of calculations performed using these “skewed” ranges. Or am I missing something?
I would expect tabular results to display the “mid-point” values and graphs to display the “mid-point” line together with a feather (in this case a broader feather downwards between 20% and the mid-point 45%, and a thinner feather upwards between the mid-point 45% and 50%)
BUT you can nicely get the featherings to display differently above and below when using min/max … although the “mid-point” line still doesn’t fall on the desired values of 45%, 45% and 50%:
The distribution charts for the 3 sample months turn out as follows - and they are not affected at all by 90% or min/max uncertainty interval settings. I would have expected at least the Range values displayed in the distribution graphs to change
