How can I normalize a histogram (dividing by the maximum number of observations in a bin, so that the histogram maximum = 1).

Hello everyone,

I need to normalize a histogram, so that the histogram maximum is equal to 1. I want to divide by the maximum No. of values in each bin.

I tried to normalize using cumcount but it doesn't give me the graph I want although it gives me the limit of histogram between 0, and 1.

I need to plot a graph like this

any suggestions??

Thanks

Ebtesam

Answer 1

[Image Analyst]

Just divide by the max

counts = histcounts(data);
counts = counts / max(counts);

Seems so obvious - am I missing something?

[Ebtesam Farid]

Hello @Image Analyst

I tried this but, this makes my x-values less than 1, doesn't give me the probability. I want to make the y-lim from (0 - 1) if I normalize my historgam by probability, maybe I wrote the question in a wrong way, am sorry, for that reason I attached the graph I want to mimic.

any suggestions?

[Image Analyst]

What did you try? You forgot to attach it. My code does not change the x values. It only changes the y values (bin heights, counts). How did you create the red and black curves? Usually histograms are bar charts and neither of those is. It looks like the black line is some kind of actual data and the red line is a log-normal, or rayleigh fit to the black data.

[Ebtesam Farid]

sorry again. first, my measurements are estimated precipitation from one of GNSS stations , and I am trying to plot the histogram for the precipitable water vapor variable to look to its shape to see if it's lognormally distributed or not, and the paper that I attached a graph from it, did what you said, plotted the histogram of varaible (black curve) then lognormally fitted it (red curve).

when I tried your codes,

counts = histcounts(data);
counts = counts / max(counts);
h = hist(data, counts);

it give me the values on x that's between 0 and 1, maybe I mistaken in something.

sorry for replying late

Answer 2

[Image Analyst]

I think you need to use kstest(), kstest2(), or lillietest() to determine if your samples are normally distributed. Of course take the log of them first if you want to see if the log of your values are normally distributed.

[Ebtesam Farid]

Thanks a lot for your suggestions