User talk:Steven J. Koch
From OpenWetWare
no, thank YOU
I love the idea of using a wiki. Even if it is used to post photos of eyes and talk about the squiggly dots, which I have yet to notice, but I will make a point of it.--Bradley A. Knockel 10:06, 22 August 2007 (EDT)
Re: comments on Physics307L:People/Mondragon/Notebook/071003#Compilation_of_plots
Have you thought of a reason for that possible 
weighting factor that you suggested may be missing in my chisquare fit? I looked at the article about Poisson distribution on wikipedia and I found what you may have been thinking. The standard deviation of a Poisson distribution with a mean λ is 
. I don't think this has much to do with the uncertainty of a count. If I were to include an uncertainty in count, that uncertainty would have more to do with how fast our apparatus could react to an event, I think. If the equipment can only count reliably as fast as x counts per millisecond, the uncertainty would rise with an increase in count number, as the , but it would also increase with a decrease in bin size (dwell time).
I might analyze how this kind of error would effect data. It might also make a good experiment if there was some way to get the detector to detect things that happen a an adjustable constant rate rather than an average rate.
Where the maybe should have come in my analysis was in that final plot. I used the error in my fits as error bars for the λs. No, I did it right, the error bars are estimation of how uncertain the measurement (or calculation) of λ. No, the error bars are probably wrong, and the correct way to calculate it is the way you mentioned Dr. Gold taught. I should have paid more attention to Dr. Gold's lectures, I doubt I have notes on how he calculated the uncertainty of a chi-square fit.
I will definitely have to analyze how close using the average of the data as λ gets to the λ calculated by a chi-square fit, since as both you, Bradley, and the world has pointed out, λ is the average of a Poisson distro. It looks like the Poisson distribution has another way to fit it not based on chi-square but still based on maximum likelihood, from the looks of the wiki article.
Looks like I have plenty to write about. --Tomas A. Mondragon 18:50, 25 October 2007 (CDT)
