Thread: Histogram

Hello,
i have a data of 10.000.000 numbers and need to get out a histogram, or better said a dataset, which gives out a histogram when directly plotted (i already took the raw data an plottet a histogram with gnuplot, was no problem, but thats not my task).
The problem is, that im very new to C, and i couldnt find any good tutorials or sides in google.
The histogram should be on logscale, the x axis has a range from 0.01-150 and it should be devided into bits which contains at least 10 numbers. So the width of the bits would not be constant.
i would be thankful for any help, tips, links or part of codes.

okay, thank you. Maybe i will need help when i arrived at step 3 :-).

Hello again. Now i am that far, that i get a nice histogram. But until now i have chosen a concrete binwidth. Now i want to have dynamical bins, meaning if the number of data in a bin is under some treshhold, for expample 10, the binwidth should increase.
I would be thankful for every hints how to construct the right array, links or anything.

Thank you, but your answer does not help me or i did not understand you. i already made an histogram, i found the max- and minvalues, i chose 100 for N ( had a larger sample ofdata). But i want to get a minimum number of numbers in 1 bin. Take your example: if you would subdivide the data into only 10 bins of a size of 0.01-2, etc... (ignoring the 3 values higher than 20), almost all of your data woud be in bin 1 the other bins would have one or zero numbers in it.

my idea is: if the number (j) <4 let the arraysize expand until 4 is reached...
(i know that i will have just brider bins with the same hight for some data, but i will devide the frequency by the binwidth and give that out later.
(sorry for my bad english :-D)

Sorry, i meant binwidth instead of "arraysize"

With problems like this, I've found it VERY helpful to take a VERY small example of the problem, and solve it with paper and pencil - (or a sample program I code up), and note the pattern I used to solve it. You may need to repeat it a few times to see it, but when you do see it, write down the steps you took to solve it. That is the "backbone" of your pseudo code right there, for your program.

Break that "backbone" down into smaller steps suitable for the computer, and now you have your pseudo code for your program. Change that into working code, function by function, putting off the details until later in the process.

Just off the cuff, I'm thinking of a a nested set of loops in a function, where the number of bins in total has been passed to it (and that number will change each time it's passed). Given that number of bins, the function works through the data, trying different bin widths, and recording the one most equal in distribution of data per bin.
If the distribution goes way off track, the inner loop of the function breaks, and another width is tried. It would be a slow process, most tedious to do by hand with any large amount of data, but the computer would romp through it OK.

Can you make a small problem set for an example?

Nevermind i fixed the problem myself. Once you get the idea its always pretty easy. But thanks for answering