What is a One-Sided Normal Distribution?
What is a One-Sided Specification?
Hmmm I have never heard of a one sided normal distribution. There are one or two tail tests.. maybe that is what you mean?
One sided specification means that you have a failure limit on the high or low side but not on both.
I guess it could be for a “barrier” type of feature, such as run-out on a shaft, flatness, etc. Since 0 is the best it could be, there is only an “upper” tolerance. This might answer both the distribution and the tolerance question.
A one sided normal distribution is better referred to as a “half-normal”
distribution and consists of only the right or left h
RickManCoffee:A one sided normal distribution is better referred to as a
“half-normal” distribution and consists of only the right or left half of the
bell shaped curve. It is observed and used when the parameter of interest has a
natural upper or lower limit. Other cases are when only the absolute value of
the measurement is used. I have seen it used while measuring:- Hole locations – deviations from center- Technical efficiency – where 100% is the maximum.- Wavelength correlation for matching spectra – where
maximum correlation is 1.0- Deviations from standard operating temperature where
previous data showed only positive deviations.- Contact resistance of an electrical bonding process –
where the target is zero resistance. In order to calculate a process capability you can either:- Transform the data to a more normal (or at least
symmetric) distribution and use a single specification limit using the normal
distribution.- Assume a different distribution that has a natural limit
on one or both sides. For example, assume a log-normal or binomial.- Calculate the process capability assuming and using only
the half of the distribution that is applicable. I don’t know of any software
that will allow this; you’ll have to do it by hand. Neither assumption
allows for the discussion or quantification of “centering”. Cheers, BTDT
The forum ‘General’ is closed to new topics and replies.