Bin size and number of bins

[JoeMichael57]

I have been very impressed with this software and have been using it to understand differences between EBSD maps from a variety of samples. To make close comparisons I would like to have the same number of bins and bin sizes for each analysis. I can set the number of bins but I do not know how to set the bin size. Is this possible ?

Thanks
Joe Michael

[marcoalopez]

Hi Joe. Thank you for using GrainSizeTools.

To generate a histogram with a specific bin size you can use the np.arange() function as follows:

# import numpy if you have not already done so
import numpy as np

# Create an array containing the limits of the bins using np.arange(Start of interval, End of interval, Spacing between values)
bin_size = 5.0
bin_edges= np.arange(dataset['diameters'].min(), dataset['diameters'].max() + bin_size, bin_size)

# generate the histogram
fig, ax = plot.distribution(dataset['diameters'], binsize=bin_edges)

In the example above, I have used the minimum and maximum values of the distribution as the start and end values, and a step size of 5.0, but you can use any value you like. Note that the bin size should also be added to the maximum value to ensure no data is left out of the histogram.

Let me know if it works for you.

Best, Marco.

PS. The way plot.distribution is implemented, if you pass a number, e.g. binsize=5.0, the script will calculate the closest bin size based on your data range, but not the exact bin size you wanted. I think I will re-implement this in the next version (3.3) so that the user can easily choose whether to specify the bin size or the number of bins. I will also add this to the documentation/wiki asap. Thank you for raising this issue

[JoeMichael57]

Thank you for the response. I should have been more specific in my question. How do I implement your response with the stereology module? If I have two lists of equivalent circle diameters and I want to have grain size distributions calculated by the Saltykov method with the same number of bins and bin sizes so that I can directly compare them on the same plot, can I do this? I understand that there may be empty bins in some cases, but I am trying to make a point about the number of grains needed for these methods to work well and what happens if too few grains are used.

Joe

[marcoalopez]

Ok, this is a different problem.

In the case of Saltykov's method, you can only set the number of bins/classes by design, not the bin size. If you have more than one population of grain sizes, the bin size will only match if the range of grain sizes is similar between the populations, which I assume will not be the case. One option is to set an ad hoc upper limit for all your distributions. You cannot do this directly, as the method is not designed to do so, but you can modify a line of code to do it. In particular, if you open the stereology.py archive and go to lines 107-109 you will see this:

freq, bin_edges = np.histogram(
        diameters, bins=numbins, range=(minimo, diameters.max()), density=True
    )

On line 108, where it specifies range=(minimo, diameters.max()), you can replace diameters.max() with the upper limit you want, and all histograms will look the same as long as you use the same number of classes.

Another way of making comparisons between populations is to set a predefined range (or ranges) of grain sizes and then calculate the volume fraction and compare between populations. This can be done directly with the Saltykov function or, if you use the two-step method, with the calc_volume_fraction function.