univgui.m

常用ROBUST STATISTICAL

function univgui(arg)
% UNIVGUI Distribution Shapes - Univariate
%
% This GUI function allows one to explore the univariate distributions in
% the data set. These are the distributions of the columns of the data
% matrix X.
%
% One can call it from the edagui GUI or stand-alone from the command
% line. To call from the command line use
%
%       univgui
%
%   Exploratory Data Analysis Toolbox, April 2005
%   Martinez and Martinez, Exploratory Data Analysis with MATLAB
%   CRC Press

% First set up the layout if it does not exist.
flg = findobj('tag','univgui');
if isempty(flg)
    % then create the gui
    dsulayout
elseif nargin == 0
    % bring it forward
    figure(flg)
end

if nargin == 0
    arg = ' ';
end
if strcmp(arg,'dispbox')
    % Display the side-by-side boxplots.
    dispbox
elseif strcmp(arg,'disphist')
    % Display the histograms of the columns. These will be displayed in a
    % matrix-like layout.
    disphist
elseif strcmp(arg,'dispqq')
    % Display q-q plots of the columns of the data matrix. These will be
    % displayed in a matrix-like layout.
    dispqq
elseif strcmp(arg,'close')
    % in other gui's we will do some housekeeping. With this gui, we do not
    % have to do so. Obviously, the user will want to keep the data from
    % the loadgui for other applications. That is the purpose.
    tg = findobj('tag','univgui');
    H = get(tg,'userdata');
    if ~isempty(H.plots)
        button = questdlg('Closing this GUI will close the associated plot windows.',...
            'Closing GUI Warning','OK','Cancel','Cancel');
        if strcmp(button,'Cancel')
            return
        else
            close(H.plots)
        end
    end
    delete(tg)
end

%%%%%%%%%%%%%%%%%%%%%%%%%  SUB FUNCTIONs %%%%%%%%%%%%%%%%%%%%%%%%%%%%
function dispbox
% This function displays the side-by-side boxplots.
% Get the data matrix. Get the GUI info.
ud = get(0,'userdata');
if isempty(ud.X)
    errordlg('You must load up some data first.')
    return
end
[n,p] = size(ud.X);
tg = findobj('tag','univgui');
H = get(tg,'userdata');
% Get the chosen type of plot.
boxchoice = get(H.popbox,'value');
% Get the dimensions to plot.
dims = get(H.dimbox,'string');
try
    if strcmp('all',lower(dims))
        % Wants to plot all dimensions
        dim = 1:p;
    else
        % Wants just a subset - convert to numbers.
        eval(['dim = [' dims '];'])
    end
catch
    errordlg('Data entry error in edit box.','Data Enty Error')
    return
end
% Now do the plots based on the choice.
switch boxchoice
    case 1
        % This is the regular boxplots
        hf = figure;
        set(hf,'numbertitle','off','name','EDA: Boxplots')
        % Upon figure close, this should delete from the array.
        set(hf,'CloseRequestFcn',...
            'tg = findobj(''tag'',''univgui''); H = get(tg,''userdata''); H.plots(find(H.plots == gcf)) = []; set(tg,''userdata'',H); delete(gcf)')
        H.plots = [H.plots, hf];
        set(tg,'userdata',H)
        boxplot(ud.X(:,dim));
    case 2
        % This is the nothed boxplots
        hf = figure;
        set(hf,'numbertitle','off','name','EDA: Notched Boxplots')
        % Upon figure close, this should delete from the array.
        set(hf,'CloseRequestFcn',...
            'tg = findobj(''tag'',''univgui''); H = get(tg,''userdata''); H.plots(find(H.plots == gcf)) = []; set(tg,''userdata'',H); delete(gcf)')
        H.plots = [H.plots, hf];
        set(tg,'userdata',H)
        boxplot(ud.X(:,dim),1);
    case 3
        % Box-percentile plot
        hf = boxprct(ud.X(:,dim));
        set(hf,'numbertitle','off','name','EDA: Box-Percentile Plot')
        % Upon figure close, this should delete from the array.
        set(hf,'CloseRequestFcn',...
            'tg = findobj(''tag'',''univgui''); H = get(tg,''userdata''); H.plots(find(H.plots == gcf)) = []; set(tg,''userdata'',H); delete(gcf)')
        H.plots = [H.plots, hf];
        set(tg,'userdata',H)
    case 4
        % Histplot
        hf = boxp(ud.X(:,dim),'hp');
        set(hf,'numbertitle','off','name','EDA: Histplots')
        % Upon figure close, this should delete from the array.
        set(hf,'CloseRequestFcn',...
            'tg = findobj(''tag'',''univgui''); H = get(tg,''userdata''); H.plots(find(H.plots == gcf)) = []; set(tg,''userdata'',H); delete(gcf)')
        H.plots = [H.plots, hf];
        set(tg,'userdata',H)
end
        
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%5
function disphist
% This function displays the histograms.
% The plots for this choice will depend on whether the user chooses to have
% plot all of the dimensions (shown in a plotmatrix format) or one
% dimension (shown in one axis only).
% 
% Get the data matrix. Get the GUI info.
ud = get(0,'userdata');
if isempty(ud.X)
    errordlg('You must load up some data first.')
    return
end
[n,p] = size(ud.X);
tg = findobj('tag','univgui');
H = get(tg,'userdata');
% Get the chosen type of plot.
histchoice = get(H.pophist,'value');
% Get the dimensions to plot.
dims = get(H.dimhist,'string');
try
    if strcmp('all',dims)
        % Wants to plot all dimensions
        dim = 1:p;
    else
        % Wants just a subset - convert to numbers.
        eval(['dim = [' dims '];'])
    end
catch
    errordlg('Data entry error in edit box.','Data Enty Error')
    return
end
% Now do the plots based on the choice.
pp = length(dim);
% Get the layout of the subplots.
nr = round(sqrt(pp));
nc = ceil(pp/nr);
% Get the bandwidths. Will be a vector of bandwidths - based on number
% wanting to plot.
switch histchoice
    case 1
        % Use Normal reference rule (really Scott's rule)
        sig = std(ud.X(:,dim));
        bw = 3.5*sig*n^(-1/3);
        hf = figure;
        set(hf,'numbertitle','off','name','EDA: Probability Density Histogram - Normal Reference Rule')
        % Upon figure close, this should delete from the array.
        set(hf,'CloseRequestFcn',...
            'tg = findobj(''tag'',''univgui''); H = get(tg,''userdata''); H.plots(find(H.plots == gcf)) = []; set(tg,''userdata'',H); delete(gcf)')
        H.plots = [H.plots, hf];
        set(tg,'userdata',H)
    case 2
        % Freedman-Diaconis Rule
        for i = 1:pp
            ti = dim(i);
            q = quartiles(ud.X(:,ti));
            IQR = q(3) - q(1);
            bw(i) = 2*IQR*n^(-1/3);
        end
        hf = figure;
        set(hf,'numbertitle','off','name','EDA: Probability Density Histogram - Freedman/Diaconis Rule')
        % Upon figure close, this should delete from the array.
        set(hf,'CloseRequestFcn',...
            'tg = findobj(''tag'',''univgui''); H = get(tg,''userdata''); H.plots(find(H.plots == gcf)) = []; set(tg,''userdata'',H); delete(gcf)')
        H.plots = [H.plots, hf];
        set(tg,'userdata',H)
    case 3
        % Sturge's Rule
        k = round(1 + log2(n));
        rng = max(ud.X(:,dim)) - min(ud.X(:,dim));
        bw = rng/k;
        hf = figure;
        set(hf,'numbertitle','off','name','EDA: Probability Density Histogram - Sturges Rule')
        % Upon figure close, this should delete from the array.
        set(hf,'CloseRequestFcn',...
            'tg = findobj(''tag'',''univgui''); H = get(tg,''userdata''); H.plots(find(H.plots == gcf)) = []; set(tg,''userdata'',H); delete(gcf)')
        H.plots = [H.plots, hf];
        set(tg,'userdata',H)
        
end
% Now that we have the bandwidths - do the histogram and plots.
for i = 1:pp
    subplot(nr,nc,i);
    ti = dim(i);
    % Am going to cheat and convert bandwidth to number of bins. Then use
    % the hist function. Err on the side of too many bins.
    rng = max(ud.X(:,ti)) - min(ud.X(:,ti));
    k = ceil(rng/bw(i));
    [nuk,xk] = hist(ud.X(:,ti),k);
    h = xk(2) - xk(1);
    bar(xk,nuk/(n*h), 1, 'w')
    if isnumeric(ud.varlab)
        title(num2str(ud.varlab(ti)));
    elseif iscell(ud.varlab)
        title(ud.varlab{ti});
    end
end    

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
function dispqq
% This function displays the q-q plots
% Get the data matrix. Get the GUI info.
ud = get(0,'userdata');
if isempty(ud.X)
    errordlg('You must load up some data first.')
    return
end
[n,p] = size(ud.X);
tg = findobj('tag','univgui');
H = get(tg,'userdata');
% Get the chosen type of plot.
qqchoice = get(H.popqqt,'value');
% Get the dimensions to plot.
dims = get(H.dimqq,'string');
try
    if strcmp('all',dims)
        % Wants to plot all dimensions
        dim = 1:p;
    else
        % Wants just a subset - convert to numbers.
        eval(['dim = [' dims '];'])
    end
catch
    errordlg('Data entry error in edit box.','Data Enty Error')
    return
end
% Now do the plots based on the choice.
pp = length(dim);
% Get the layout of the subplots.
nr = round(sqrt(pp));
nc = ceil(pp/nr);
% Get the bandwidths. Will be a vector of bandwidths - based on number
% wanting to plot.
% This will generate the random variables to plot against.
% This will be a matrix of values. 
randmat = zeros(n,pp);

switch qqchoice
    case 1
        % Normal distribution
        randmat = randn(n,pp);
        hf = figure;
        set(hf,'numbertitle','off','name','EDA: QQ Plot - Normal Distribution')
        % Upon figure close, this should delete from the array.
        set(hf,'CloseRequestFcn',...
            'tg = findobj(''tag'',''univgui''); H = get(tg,''userdata''); H.plots(find(H.plots == gcf)) = []; set(tg,''userdata'',H); delete(gcf)')
        H.plots = [H.plots, hf];
        set(tg,'userdata',H)
    case 2
        % Exponential distribution
        % First estimate the parameters. Use those in the rng function.
        parmhat = expfit(ud.X(:,dim));
        for i = 1:pp
            ti = dim(i);
            randmat(:,i) = exprnd(parmhat(i),n,1);
        end
        hf = figure;
        set(hf,'numbertitle','off','name','EDA: QQ Plot - Exponential Distribution')
        % Upon figure close, this should delete from the array.
        set(hf,'CloseRequestFcn',...
            'tg = findobj(''tag'',''univgui''); H = get(tg,''userdata''); H.plots(find(H.plots == gcf)) = []; set(tg,''userdata'',H); delete(gcf)')
        H.plots = [H.plots, hf];
        set(tg,'userdata',H)
    case 3
        % Gamma distribution
        for i = 1:pp
            ti = dim(i);
            phat = gamfit(ud.X(:,ti));
            randmat(:,i) = gamrnd(phat(1),phat(2),n,1);
        end
        hf = figure;
        set(hf,'numbertitle','off','name','EDA: QQ Plot - Gamma Distribution')
        % Upon figure close, this should delete from the array.
        set(hf,'CloseRequestFcn',...
            'tg = findobj(''tag'',''univgui''); H = get(tg,''userdata''); H.plots(find(H.plots == gcf)) = []; set(tg,''userdata'',H); delete(gcf)')
        H.plots = [H.plots, hf];
        set(tg,'userdata',H)
    case 4
        % Chi-square distribution
        randmat = chi2rnd(n-1,n,pp);
        hf = figure;
        set(hf,'numbertitle','off','name','EDA: QQ Plot - Chi-Square Distribution')
        % Upon figure close, this should delete from the array.
        set(hf,'CloseRequestFcn',...
            'tg = findobj(''tag'',''univgui''); H = get(tg,''userdata''); H.plots(find(H.plots == gcf)) = []; set(tg,''userdata'',H); delete(gcf)')
        H.plots = [H.plots, hf];
        set(tg,'userdata',H)
    case 5
        % Lognormal distribution
        for i = 1:pp
            ti = dim(i);
            phat = lognfit(ud.X(:,ti));
            randmat(:,i) = lognrnd(phat(1),phat(2),n,1);
        end
        hf = figure;
        set(hf,'numbertitle','off','name','EDA: QQ Plot - Lognormal Distribution')
        % Upon figure close, this should delete from the array.
        set(hf,'CloseRequestFcn',...
            'tg = findobj(''tag'',''univgui''); H = get(tg,''userdata''); H.plots(find(H.plots == gcf)) = []; set(tg,''userdata'',H); delete(gcf)')
        H.plots = [H.plots, hf];
        set(tg,'userdata',H)
    case 6
        % Uniform distribution
        for i = 1:pp
            ti = dim(i);