Re: weighted variance comparisons

On Aug 30, 4:49 am, chris <hunte...@xxxxxxxxxxxxxxx> wrote:
Thanks for all your comments. I have solved my immediate program by
putting the raw data into Access using it to overcome the limitations
of Excel regarding maximum number of records.

I have been able to calculate the weighted variances using the formula
you gave and checked it with using weights of 1 so I am happy with my
results.

The methodology was a bit clunky but it worked. Now that I have
finished this project and have some time, I will investigate R, it
sounds worthwhile.

I am still unsure of how to treat frequency data. The formula given by
peterp does not seem to me to take any account of the variance of the
data within each weighting bin. ie the variance of the data for the
6.25 samples could be quite different from the variance of the data
within the 12.5 bin. How is this taken into account?

Thanks again
Chris

Unless you know what the variance within each bin is, you cannot
recover that information from the between bin variability. The total
variability in the values can be decomposed into the within-bin and
the between-bin variance. This is the same idea as in the analysis of
variance. So, to calculate the total variance, you not only need the
mean of each bin, but also the amount of variance within each bin.

m00es

.

Relevant Pages

  • Re: uC selection
    ... the variance changes by 3:1 as well as the mean. ... than about 1/10th of a bin with simple methods. ... narrow channels is going to be 3 times higher than in the wide channels ... To plot the histogram, you divide the count in each channel by its ...
    (sci.electronics.design)
  • Re: uC selection
    ... the variance changes by 3:1 as well as the mean. ... than about 1/10th of a bin with simple methods. ... narrow channels is going to be 3 times higher than in the wide channels ...
    (sci.electronics.design)
  • Re: uC selection
    ... the variance changes by 3:1 as well as the mean. ... than about 1/10th of a bin with simple methods. ... In the case where I used method 1 I was actually making a ADC good to ... (assuming a 0.5 LSB DNL spec). ...
    (sci.electronics.design)
  • Re: uC selection
    ... On Mar 16, 10:02 pm, Phil Hobbs ... the variance changes by 3:1 as well as the mean. ... than about 1/10th of a bin with simple methods. ... narrow channels is going to be 3 times higher than in the wide channels ...
    (sci.electronics.design)
  • Re: About Bootstrap and replacement
    ... It's also not too difficult to show that the variance of your estimate of the mean is larger than the standard estimate, so your estimator will be worse than the bootstrap estimator. ... algorithm without replacement and my *strange weighting* are ... If you had bothered to do the sums, hten you would have found out that the variance of a weighted mean of x_i with weights w_i is * sum w_i^2)/^2. ... From this you can work out that the minimum varaince occurs when the weights are equal. ...
    (sci.stat.math)