Dear EJMR friends
How do I get a point estimate for standard deviation when I know sample size, sample mean, sample max, and sample min?
Arigato
stats question
-
Forgot to mention: if you have some notion of the distribution (other than normal), you could do some maximum likelihood using the mean and order statistics to estimate the standard deviation. But, if you are OK with a normality assumption, Parkinson is a nice, simple estimator.
Also forgot to note that I remain the - Q36spacemodulator
-
Q36spacemodulator and other EJMR friends,
I read Parkinson's paper carefully and it does not apply to my case. Parkinson assumes that the diffusion process is monitored continuously over time so that the sample max and the sample min are actually the max and the min along the realized path.
This is not my case. In my case the process is discretely observed (say, at full hours like 9:00, 10:00, 11:00 etc), someone recorded the max, min, and mean of those say N=24 hours, and then lost the individual hourly observations and just kept the max, min, and mean.
How do I estimate the standard deviation in this case?
Domo arigato