%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Script 4 Working with cell arrays
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
rand('state',37); % set arbitrary seed for uniform draws
randn('state',37); % set arbitrary seed for normal draws

m1=reshape(repmat((1:1:8),4,1),32,1); %design first column
n=length(m1);
m2=2+1.4*randn(n,1); %design second column
m3=randsample([0 1],n,true)'; %this samples randomly n elements from the 
% set [0 1] with replacement (drop "true" if you ever want to sample from a
% set of mubers w/o replacement)
M=[m1 m2 m3];
k=size(M,2);

MC=mat2cell(M,4*ones(8,1),k);
mc1=MC{1}; %inspect contents of first cell
mc1c2=MC{1}(:,2); %get second column of first cell

%Ex.1:
MC1=MC; %to preserve the original MC
for i=1:8
    MC1{i}=sortrows(MC{i},3); %means sort the rows in MC{i} in order 
    %of the elements in the third column
end
MC1=cell2mat(MC1);

%Ex.2:
Msum=zeros(8,1); %pre-allocate a zero vector to collect results
for i=1:8
    Msum(i)=sum(MC{i}(:,2));
end

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
% Working with unbalanced panels
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
load c:\mlab_tut\from_stata.raw;
data = from_stata;
clear from_stata;

% describe contents of data
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
%1  runid           running observation id  
%2  orig_solid      original solicitor ID
%3  amount          contributions (2005 dollars)
%4  gender          solicitor Gender (1=male)
%5  race            solicitor Race (1=white)
%6  height          solicitor height (inches)
%7  weight          solicitor weight (lbs)
%8  beauty          beauty index 
%9  spunk           sum of personality scores
%10 gen_resp        observed gender of respondent (1=male)                                               
%11 raceresp        observed race of respondent(1=white)                                           
%12 avg_HHsize      average household size (Census block)
%13 inc000          Census tract median income (2005 dollars)                
%14 lottery         fundraising package included lottery component

data=sortrows(data,2);

y = data(:,3); 
X = data(:,4:end); 
n=length(y);
k=size(X,2);

% get vector with panel sizes
sol_id=data(:,2);
sol_vec=[]; %assume we don't even know how many individuals there are, so 
% instead of pre-allocating a vector with a specific number of zeros we
% simply pre-allocate an empty vector;

j=0;
for i=2:n
    if sol_id(i)~=sol_id(i-1)% check if subsequent elements in sol_id are 
        % equal or not
        j=i-1-sum(sol_vec); %j will give the panel size
        sol_vec=[sol_vec;j];
    end
end
sol_vec=[sol_vec;n-sum(sol_vec)]; % tag on size of last panel

% make sure this worked:
test=mat2cell(sol_id,sol_vec,1);
test{1};
test{32};
test{end};

% now partition X into cells
XC=mat2cell(X,sol_vec,k);


%%%%%%%%%%%%%
% Practice
%%%%%%%%%%%%%%%
%Ex 1:
a=[0 3 5 0 3 5 0 3 5 0 3 5]';
a=sort(a);
A=mat2cell(a,4*ones(3,1),1)

%Ex2:
for i=1:3
    A{i}=A{i}';
end
AA=cell2mat(A);

%equivalently:
AA=(reshape(a,4,3))';


%Ex3:
rand('state',37); % set arbitrary seed for uniform draws
randn('state',37); % set arbitrary seed for normal draws
g1=randsample([1:1:5],30,true)';
g2=ones(30,1);
g3=8*rand(30,1);
g4=randn(30,1);
G=[g1 g2 g3 g4];
n=length(G);
k=size(G,2);

G=sortrows(G,1);
id=G(:,1);
s=[]; 
j=0;
for i=2:n
    if id(i)~=id(i-1)% check if subsequent elements in sol_id are 
        % equal or not
        j=i-1-sum(s); %j will give the panel size
        s=[s;j];
    end
end
s=[s;n-sum(s)]; % tag on size of last panel
GC=mat2cell(G,s,k);

sl=length(s);
for i=1:sl
    GC{i}=sortrows(GC{i},4);
end

GC=cell2mat(GC);