This function reads the dataset and determines the location of the missing observations (defined by NaNs)
0001 function [i,n,s,j] = describe_missing_data(data) 0002 % This function reads the dataset and determines the location of the missing observations (defined by NaNs) 0003 0004 %@info: 0005 %! @deftypefn {Function File} {[@var{i}, @var{n}, @var{s}, @var{j} ] =} describe_missing_data (@var{data}, @var{gend}, @var{nvarobs}) 0006 %! This function reads the dataset and determines where are the missing observations. 0007 %! 0008 %! @strong{Inputs} 0009 %! @table @var 0010 %! @item data 0011 %! Real matrix (T-by-N) for the dataset. 0012 %! @end table 0013 %! 0014 %! @strong{Outputs} 0015 %! @table @var 0016 %! @item i 0017 %! cell array (1-by-T). Each element is a @math{p_t\times 1} column vector of indices targeting the non-NaN variables at time t. 0018 %! @item n 0019 %! Integer scalar. The effective number of observations: 0020 %! @math(n=\sum_{t=1}^T p_t) 0021 %! @item s 0022 %! Integer scalar. The value of the time index such that @math(p_t=p_s) for all @math(t\geq s). 0023 %! @item j 0024 %! cell array (1-by-N). Each element is a column vector targeting to the non-NaN observations of a variable. 0025 %! @end table 0026 %! 0027 %! @end deftypefn 0028 %@eod: 0029 0030 % Copyright (C) 2008-2011 Dynare Team 0031 % stephane DOT adjemian AT univ DASH lemans DOT fr 0032 % 0033 % This file is part of Dynare. 0034 % 0035 % Dynare is free software: you can redistribute it and/or modify 0036 % it under the terms of the GNU General Public License as published by 0037 % the Free Software Foundation, either version 3 of the License, or 0038 % (at your option) any later version. 0039 % 0040 % Dynare is distributed in the hope that it will be useful, 0041 % but WITHOUT ANY WARRANTY; without even the implied warranty of 0042 % MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 0043 % GNU General Public License for more details. 0044 % 0045 % You should have received a copy of the GNU General Public License 0046 % along with Dynare. If not, see <http://www.gnu.org/licenses/>. 0047 0048 [observation_index,variable_index] = find(~isnan(data)); 0049 [T,N] = size(data); 0050 0051 i = cell(1,T); 0052 j = cell(1,N); 0053 missing_observations_counter = NaN(T,1); 0054 0055 for obs=1:T 0056 idx = find(observation_index==obs); 0057 tmp = variable_index(idx); 0058 missing_observations_counter(obs,1) = N-length(tmp); 0059 if rows(tmp(:)) 0060 i(obs) = { tmp(:) }; 0061 else 0062 i(obs) = { [] }; 0063 end 0064 end 0065 0066 missing_observations_counter = cumsum(missing_observations_counter); 0067 0068 n = length(variable_index); 0069 0070 if ~missing_observations_counter 0071 s = 1; 0072 else 0073 tmp = find(missing_observations_counter>=(T*N-n)); 0074 s = tmp(1)+1; 0075 end 0076 0077 if nargout>3 0078 for var=1:N 0079 idx = find(variable_index==var); 0080 tmp = observation_index(idx); 0081 j(var) = { tmp(:) }; 0082 end 0083 end 0084 0085 0086 %@test:1 0087 %$ % Define a data set. 0088 %$ A = [ 1 1 ; ... 0089 %$ 1 NaN ; ... 0090 %$ NaN 1 ; ... 0091 %$ 1 1 ; ... 0092 %$ NaN NaN ; ... 0093 %$ 1 NaN ; ... 0094 %$ 1 NaN ; ... 0095 %$ 1 1 ; ... 0096 %$ 1 1 ; ... 0097 %$ 1 1 ; ... 0098 %$ 1 1 ]; 0099 %$ 0100 %$ % Define expected results. 0101 %$ eB = cell(1,11); 0102 %$ eB(1) = { transpose(1:2) }; 0103 %$ eB(2) = { 1 }; 0104 %$ eB(3) = { 2 }; 0105 %$ eB(4) = { transpose(1:2)}; 0106 %$ eB(5) = { [] }; 0107 %$ eB(6) = { 1 }; 0108 %$ eB(7) = { 1 }; 0109 %$ eB(8) = { transpose(1:2) }; 0110 %$ eB(9) = { transpose(1:2) }; 0111 %$ eB(10) = { transpose(1:2) }; 0112 %$ eB(11) = { transpose(1:2) }; 0113 %$ eC = 16; 0114 %$ eD = 8; 0115 %$ eE = cell(1,2); 0116 %$ eE(1) = { [1; 2; 4; transpose(6:11)] }; 0117 %$ eE(2) = { [1; 3; 4; transpose(8:11)] }; 0118 %$ 0119 %$ % Call the tested routine. 0120 %$ [B,C,D,E] = describe_missing_data(transpose(A)); 0121 %$ 0122 %$ % Check the results. 0123 %$ t(1) = dyn_assert(B,eB); 0124 %$ t(2) = dyn_assert(C,eC); 0125 %$ t(3) = dyn_assert(D,eD); 0126 %$ t(4) = dyn_assert(E,eE); 0127 %$ T = all(t); 0128 %@eof:1