Re: How do I claim a measurement is Gaussian distribution?


mailcwc@xxxxxxxxx wrote:
Hi, all

I am a student from engineering. I have never taken a statistics
before, but I do have background of probability.
My current research need to claim a measurement is Gaussian
distributed. What is the standard method for this?

A measurement? You can never tell.

A group of measurements assumed to have come from the same
normal distribution, then you have the standard problem of validation
of normality based on the observed data.

The standard, and by far the most effective method, is to do a Q-Q
plot of the empiricial cdf of the data against the cdf of a theoretical
normal distribution. See, e.g.,

http://en.wikipedia.org/wiki/Q-Q_plot
http://www.itl.nist.gov/div898/handbook/eda/section3/qqplot.htm

I know this can be a book, please give me some keyword that I can find
in the index.
My current concern are the following questions:

1. How many samples are enough?

One. (Many samples from the same population is still one sample).

2. Is there a thumb of rule to detemine the spacing of the histogram?

Yes, many. They are ALL bad.

I think wider spacing can wipe out the noise.

And wipe out the characteristic of the normal distribution as well.

3. There are so many tests in the general statistics books. Which one
should I choose? Which godness of fit is the most important?

None is any good. Do the visual test using a Q-Q plot. It takes
experience to tell when a departure is severe vs when it is not and
the data can reasonably be considered normal.

-- Reef Fish Bob.

.

Relevant Pages

  • Re: mean and standard deviation of left-truncated normal distribution
    ... I would like to know the mean and standard deviation of a truncated ... > normal distribution, and the truncation point k. ...
    (sci.stat.math)
  • Re: random number generation
    ... Thus spake <Redmer van Tijum>: ... >> chosen skewness or kurtosis. ... >> numbers with given mean m and standard deviation sd use: ... >> The skewness of a normal distribution is 0. ...
    (comp.soft-sys.matlab)
  • mean and standard deviation of left-truncated normal distribution
    ... calculated from the mean and sd of the underlying ... normal distribution, and the truncation point k. ... I'm not 100% sure how to convert these into mean and standard ...
    (sci.stat.math)
  • Re: random numbers
    ... (mean=162.7, standard deviation=6.2) ... For a dynamic worksheet function approach, ... Both NORMINV and RAND are numerically improved in Excel 2003 (and NORMINV is ...
    (microsoft.public.excel.misc)