Specifying bin edges when fitting a normal distribution using histfit - matlab

Question

I want to fit a histogram to some data using predefined bins. All my data points are between 1 and 10, so I want the bins to start from xmin=1, and end at xmax=10, with a step of 0.5.
I use the following commands:
x = d1.data(:,4); % x is my data
H = histfit(x,10,'normal'); % fits a histogram using 10 bins
However when doing the above, bins are determined automatically per dataset and do not correspond to the edges I want. How can I ensure that the same bin edges are used for all datasets?

Answer 1

If you have access to the Curve Fitting Toolbox, I would suggest another approach that provides the required flexibility. This involves doing the fit "yourself" instead of relying on histfit:
% Generate some data:
rng(66221105) % set random seed, for reproducibility
REAL_SIG = 1.95;
REAL_MU = 5.5;
X = randn(200,1)*REAL_SIG + REAL_MU;
% Define the bin edges you want
EDGES = 1:0.5:10;
% Bin the data according to the predefined edges:
Y = histcounts(X, EDGES);
% Fit a normal distribution using the curve fitting tool:
binCenters = conv(EDGES, [0.5, 0.5], 'valid'); % moving average
[xData, yData] = prepareCurveData( binCenters, Y );
ft = fittype( 'gauss1' );
fitresult = fit( xData, yData, ft );
disp(fitresult); % optional
% Plot fit with data (optional)
figure();
histogram(X, EDGES); hold on; grid on;
plot(fitresult);
Which yields the following plot:
and the fitted model:
General model Gauss1:
fitresult(x) = a1*exp(-((x-b1)/c1)^2)
Coefficients (with 95% confidence bounds):
a1 = 19.65 (17.62, 21.68)
b1 = 5.15 (4.899, 5.401)
c1 = 2.971 (2.595, 3.348)