Am Montag, 27. Mai 2002 00:07 schrieb Joel Hammer: Ok, I can see now what you want - thought about replying off list, but I have seen more devious discussions on this list -;)
Although I don't have the data underlying your bar graph, one can see by mere visual inspection of your plot (gauss.ps) that the variance (.475 ...) resp. standard deviation (.689..) are not properly calculated, which yields the peak in your normal curve: the smaller the variance or deviation, the higher the peak. I guess the variance to be something around 5 or 6. Can you send me your raw data? I'll try to check the calculations. Klaus BTW: I have checked now http://csep1.phy.ornl.gov/mc/node19.html You are definitely right, their formula is definitely wrong. > What I would like to do is make a Gaussian normal curve that will > superimpose itself over bar graphs showing a population distribution. > The idea is to give an immediate visual impression of how far from > "normal" the population data is given the population mean and std dev. > I haven't had success with this. I can't seen to get it right. What I > see is a much higher peak of my normal curve than what I see in my > data. I have attached a plot in fact. Here is the plot file for this. > (I use a big bash script to see this stuff up for gnuplot.) > > set key left Left > u=13.35000000000000000000 > var=.47548245614035087719 > display_v=.475 > display_u=13.350 > set label 1 "mean = 13.350 " at 15.2,.13205 right > set label 2 "std dev = .689 " at 15.2,.13205*.95 right > set label 3 "std error mean = .052 " at 15.2 ,.13205*.90 right > set label 4 "count = 172 " at 15.2,.13205*.85 right > std=.68955235924500387223 > count=172 > stderrormean=.0027 > set ylabel "Result Result Frequency" > set xlabel " " > f(x)=exp(-((x-u)**2/(2*var)))/(sqrt(2*pi*var)) > plot "/tmp/plot_data_bar" using 2:1 notitle with boxes , f(x) _______________________________________________ Linux-users mailing list - http://linux-sxs.org/mailman/listinfo/linux-users Subscribe/Unsubscribe info, Archives,and Digests are located at the above URL.