Aime F. answered • 12/05/22
Tutor
4.7
(62)
24+ years Data Science experience in academia, industry and consulting
You can get the standard deviation of a Gaussian distribution from its 75th minus its 25th percentile as
σ = (Q(0.75) – Q(0.25))/(√2(erf–1(1/2) – erf–1(–1/2))).
A Gaussian has median equal to mean, and no finite minimum or maximum so those data don't seem useful to me except to check consistency. If they're not consistent, maybe a Gaussian is not the best distribution.
NM M.
Also, to further clarify, the three data sets are distinct from one another. I was hoping that I could create 3 bell curves, using a normal distribution, and layer them atop one another to create a visual that shows how the three data sets overlap.
Report
12/05/22
Aime F.
If your histogram looks reasonably symmetric then maybe use the average of the mean and median as the mean. Otherwise, perhaps use the "asymmetric generalized normal distribution" that has 3 parameters that you can fit to your data.
Report
12/05/22
NM M.
Thanks much. I will give it a try.
Report
12/05/22
NM M.
Thank you, Aime. My mean and median are different in each of the data sets. With that, along with your description, it sounds like the normal distribution won't work?12/05/22