DBMatchValues - Match values for two different sets of variables.
It is often necessary to match (associate) variables that were computed
independently, e.g. the phase precession slope and field size of hippocampal
place cells. Typically, such data are computed over repeated experiments
using batch processing tools (see <a href="StartBatch">StartBatch</a>) and stored in databases for
subsequent group analyses. Once retrieved using <a href="DBGetValues">DBGetValues</a>, they can be
matched using this function.
USAGE
[in1,in2,out1,out2,m1,m2] = DBMatchValues(eid1,eid2,rexp)
eid1,eid2 experiment IDs (see <a href="DBGetValues">DBGetValues</a>)
rexp optional regular expression (see Example 2)
OUTPUT
in1 indices of elements in eid1 that are also in eid2
in2 indices of elements in eid2 that are also in eid1
out1 indices of elements in eid1 that are not in eid2
out2 indices of elements in eid2 that are not in eid1
m1 matches (extracted patterns) in eid1
m2 matches (extracted patterns) in eid2
EXAMPLE 1
In this example, place cells were recorded on successive days while the
animal explored a maze. Phase precession slopes and firing field sizes
were stored in a database using eids like '20120213-Maze-(1,2)' (where 1,2
corresponds to tetrode 1, cluster 2) and named 'PPSlope' and 'FieldSize',
respectively.
Get slopes and sizes:
[slopes,eid1] = DBGetValues('eid like "%Maze%" and name="PPSlope"');
[sizes,eid2] = DBGetValues('eid like "%Maze%" and name="FieldSize"');
Discard values that cannot be correlated (incomplete pairs), and reorder
so that each line in both variables corresponds to the same data:
[in1,in2] = DBMatchValues(eid1,eid2);
slopes = slopes(in1);
sizes = sizes(in2);
EXAMPLE 2
In this example, cells were recorded during both wake an sleep, and their
average firing rates should be compared across behavioral conditions.
Data were stored in a database using eids like '20120213-Maze-(1,2)' or
'20130214-Sleep-(1,2)' and named 'MeanRate'. Matching them is trickier
because their eids are not pairwise identical. Here we need to extract
the relevant portions of the eids, i.e. discard 'Sleep' or 'Maze'.
Get mean rates:
[maze,eid1] = DBGetValues('eid like "%Maze%" and name="MeanRate"');
[sleep,eid2] = DBGetValues('eid like "%Sleep%" and name="MeanRate"');
Discard values that cannot be correlated (incomplete pairs), and reorder
so that each line in both variables corresponds to the same data:
[in1,in2] = DBMatchValues(eid1,eid2,'([0-9]{8}).*([0-9]*,[0-9]*)');
maze = maze(in1);
sleep = sleep(in2);
The regular expression includes two tokens (indicated by the parentheses),
namely [0-9]{8} (eight successive occurrences of a digit between 0 and 9)
and [0-9]*,[0-9]* (any number of digits, a comma, any number of digits).
SEE
See also DBGetValues, DBGetVariables, DBAddVariable.0001 function [in1,in2,out1,out2,eid1,eid2] = DBMatchValues(eid1,eid2,rexp) 0002 0003 %DBMatchValues - Match values for two different sets of variables. 0004 % 0005 % It is often necessary to match (associate) variables that were computed 0006 % independently, e.g. the phase precession slope and field size of hippocampal 0007 % place cells. Typically, such data are computed over repeated experiments 0008 % using batch processing tools (see <a href="StartBatch">StartBatch</a>) and stored in databases for 0009 % subsequent group analyses. Once retrieved using <a href="DBGetValues">DBGetValues</a>, they can be 0010 % matched using this function. 0011 % 0012 % USAGE 0013 % 0014 % [in1,in2,out1,out2,m1,m2] = DBMatchValues(eid1,eid2,rexp) 0015 % 0016 % eid1,eid2 experiment IDs (see <a href="DBGetValues">DBGetValues</a>) 0017 % rexp optional regular expression (see Example 2) 0018 % 0019 % OUTPUT 0020 % 0021 % in1 indices of elements in eid1 that are also in eid2 0022 % in2 indices of elements in eid2 that are also in eid1 0023 % out1 indices of elements in eid1 that are not in eid2 0024 % out2 indices of elements in eid2 that are not in eid1 0025 % m1 matches (extracted patterns) in eid1 0026 % m2 matches (extracted patterns) in eid2 0027 % 0028 % EXAMPLE 1 0029 % 0030 % In this example, place cells were recorded on successive days while the 0031 % animal explored a maze. Phase precession slopes and firing field sizes 0032 % were stored in a database using eids like '20120213-Maze-(1,2)' (where 1,2 0033 % corresponds to tetrode 1, cluster 2) and named 'PPSlope' and 'FieldSize', 0034 % respectively. 0035 % 0036 % Get slopes and sizes: 0037 % 0038 % [slopes,eid1] = DBGetValues('eid like "%Maze%" and name="PPSlope"'); 0039 % [sizes,eid2] = DBGetValues('eid like "%Maze%" and name="FieldSize"'); 0040 % 0041 % Discard values that cannot be correlated (incomplete pairs), and reorder 0042 % so that each line in both variables corresponds to the same data: 0043 % 0044 % [in1,in2] = DBMatchValues(eid1,eid2); 0045 % slopes = slopes(in1); 0046 % sizes = sizes(in2); 0047 % 0048 % EXAMPLE 2 0049 % 0050 % In this example, cells were recorded during both wake an sleep, and their 0051 % average firing rates should be compared across behavioral conditions. 0052 % Data were stored in a database using eids like '20120213-Maze-(1,2)' or 0053 % '20130214-Sleep-(1,2)' and named 'MeanRate'. Matching them is trickier 0054 % because their eids are not pairwise identical. Here we need to extract 0055 % the relevant portions of the eids, i.e. discard 'Sleep' or 'Maze'. 0056 % 0057 % Get mean rates: 0058 % 0059 % [maze,eid1] = DBGetValues('eid like "%Maze%" and name="MeanRate"'); 0060 % [sleep,eid2] = DBGetValues('eid like "%Sleep%" and name="MeanRate"'); 0061 % 0062 % Discard values that cannot be correlated (incomplete pairs), and reorder 0063 % so that each line in both variables corresponds to the same data: 0064 % 0065 % [in1,in2] = DBMatchValues(eid1,eid2,'([0-9]{8}).*([0-9]*,[0-9]*)'); 0066 % maze = maze(in1); 0067 % sleep = sleep(in2); 0068 % 0069 % The regular expression includes two tokens (indicated by the parentheses), 0070 % namely [0-9]{8} (eight successive occurrences of a digit between 0 and 9) 0071 % and [0-9]*,[0-9]* (any number of digits, a comma, any number of digits). 0072 % 0073 % SEE 0074 % 0075 % See also DBGetValues, DBGetVariables, DBAddVariable. 0076 % 0077 0078 0079 % Copyright (C) 2013 by Michaël Zugaro 0080 % 0081 % This program is free software; you can redistribute it and/or modify 0082 % it under the terms of the GNU General Public License as published by 0083 % the Free Software Foundation; either version 3 of the License, or 0084 % (at your option) any later version. 0085 0086 % Transform eids if necessary 0087 if nargin >= 3, 0088 eid1 = DoRegexp(eid1,rexp); 0089 eid2 = DoRegexp(eid2,rexp); 0090 else 0091 disp([ 'Example eid: ' eid1{1}]); 0092 disp([ 'Example eid: ' eid2{1}]); 0093 end 0094 % Make sure at least one of the lists contains unique values (keys) 0095 key1 = length(eid1) == length(unique(eid1)); 0096 key2 = length(eid2) == length(unique(eid2)); 0097 if ~key1 && ~key2, 0098 error('Neither list contains unique values.'); 0099 end 0100 0101 % Check which eids in list 1 are in list 2, making sure empty eids always fail the test 0102 [~,in2] = ismember(eid1,eid2); 0103 empty = cellfun('isempty',eid1); 0104 in2(empty) = 0; 0105 % The code above returns locations (items found) intermixed with zeros (items not found) 0106 % (for details, see help for 'ismember'). Split this information into separate variables. 0107 out1 = find(in2==0); 0108 in2 = in2(in2~=0); 0109 % If values in list 2 are not unique, duplicate entries in list 1 where necessary 0110 if ~key2, 0111 in2 = find(ismember(eid2,eid2(in2))); 0112 end 0113 0114 % Check which eids in list 2 are in list 1, making sure empty eids always fail the test 0115 [~,in1] = ismember(eid2,eid1); 0116 empty = cellfun('isempty',eid2); 0117 in1(empty) = 0; 0118 % The code above returns locations (items found) intermixed with zeros (items not found) 0119 % (for details, see help for 'ismember'). Split this information into separate variables. 0120 out2 = find(in1==0); 0121 in1 = in1(in1~=0); 0122 % If values in list 1 are not unique, duplicate entries in list 2 where necessary 0123 if ~key1, 0124 in1 = find(ismember(eid1,eid1(in1))); 0125 end 0126 0127 % Helper function: transform all eids, i.e. extract tokens using the regular expression then concatenate 0128 % the tokens (eids that do not match are set to the empty string '') 0129 0130 function r = DoRegexp(s,rexp) 0131 0132 r = cellfun(@(x) regexp(x,rexp,'tokens'),s,'uniformoutput',false); 0133 empty = cellfun('isempty',r); 0134 r(empty) = {''}; 0135 r(~empty) = cellfun(@(x) horzcat(x{1}{:}),r(~empty),'uniformoutput',false); 0136 0137 disp([ 'Example pattern extraction: ' s{1} ' -> ' r{1}]);