MATLAB概率密度估计有关的函数

参数估计

mle
Maximum likelihood estimates
mle是有偏估计
fitdist
对数据进行概率分布对象拟合
histfit
具有分布拟合的直方图
fitdist和histfit是无偏估计，这两者区别在于histfit直接画出直方图。
支持的分布点这里

load carsmall;
pd = fitdist(MPG,'Normal');
xi = -3:.1:60;
f = pdf(pd,xi);
n = numel(MPG);
nbins = 20;
%% 
figure
h = histogram(MPG,'Normalization','pdf','NumBins',nbins)
hold on
% plot(xi, n*h.BinWidth*f, 'LineWidth',2)
plot(xi, f, 'LineWidth',2)
% title('Miles per Gallon')
xlabel('Miles per Gallon')
ylabel('Probability Density')
box off

%% 
figure
histfit(MPG,nbins,'Normal')
xlabel('Miles per Gallon')
ylabel('Counts')

非参数估计

ksdensity
Kernel smoothing function estimate for univariate and bivariate data

kde
Kernel density estimate for univariate data

这两者的区别在于，kde中有个权值，ksdensity可以用于二维数据。Unlike ksdensity, kde does not support boundary correction methods or data censoring.

mvksdensity
Kernel smoothing function estimate for multivariate data
mvksdensity用于高维数据。

fitdist和histfit也可以用于核密度估计。histfit直接画出直方图。

%% 
load carsmall;
%% 
pd = fitdist(MPG,'Kernel');
xi = -3:.1:60;
f = pdf(pd,xi);
n = numel(MPG);
nbins = 20;
%% 
figure
h = histogram(MPG,'Normalization','pdf','NumBins',nbins);
hold on
% plot(xi,n*h.BinWidth*f,'LineWidth',2)
plot(xi, f,'LineWidth',2)

% title('Miles per Gallon')
xlabel('Miles per Gallon')
ylabel('Probability Density')
box off

%% 
figure
histfit(MPG,nbins,'Kernel')
xlabel('Miles per Gallon')
ylabel('Count')

置信区间

paramci
Confidence intervals for probability distribution parameters