Give the program two spike data sets and one
or two time intervals and it will decide if
the counts are significantly different.
this is either with a non-parametric method
or with a sqrt transformation followed by a
t-test
Usage: [H,P,M1,M2,N1,N2] = countsig(data1,data2,T1,T2,parametric,p,quiet)
Input:
Note that all times have to be consistent. If data
is in seconds, so must be sig and t. If data is in
samples, so must sig and t. The default is seconds.
data1 - structure array of spike times (required)
data2 - structure array of spike times (required)
T1 - time interval (default all)
T2 - time interval (default T1)
parametric - 0 = non-parametric (Wilcoxon)
- 1 = ttest on sqrt of counts
- 2 = Poisson assumption
(default = 0)
p - significance level (0.05)
quiet - 1 = no display 0 = display
Output:
H - 1 if different 0 if not
P - prob of result if same
M1 - mean count for data1
M2 - mean count for data2
N1 - counts for data1
N2 - counts for data20001 function[H,P,M1,M2,N1,N2] = countsig(data1,data2,T1,T2,parametric,p,quiet) 0002 % Give the program two spike data sets and one 0003 % or two time intervals and it will decide if 0004 % the counts are significantly different. 0005 % this is either with a non-parametric method 0006 % or with a sqrt transformation followed by a 0007 % t-test 0008 % Usage: [H,P,M1,M2,N1,N2] = countsig(data1,data2,T1,T2,parametric,p,quiet) 0009 % 0010 % Input: 0011 % Note that all times have to be consistent. If data 0012 % is in seconds, so must be sig and t. If data is in 0013 % samples, so must sig and t. The default is seconds. 0014 % 0015 % data1 - structure array of spike times (required) 0016 % data2 - structure array of spike times (required) 0017 % T1 - time interval (default all) 0018 % T2 - time interval (default T1) 0019 % parametric - 0 = non-parametric (Wilcoxon) 0020 % - 1 = ttest on sqrt of counts 0021 % - 2 = Poisson assumption 0022 % (default = 0) 0023 % p - significance level (0.05) 0024 % quiet - 1 = no display 0 = display 0025 % 0026 % Output: 0027 % 0028 % H - 1 if different 0 if not 0029 % P - prob of result if same 0030 % M1 - mean count for data1 0031 % M2 - mean count for data2 0032 % N1 - counts for data1 0033 % N2 - counts for data2 0034 0035 0036 if nargin < 2;error('I need 2 sets of spike data');end 0037 data1=padNaN(data1); % create a zero padded data matrix from input structural array 0038 data2=padNaN(data2); % create a zero padded data matrix from input structural array 0039 data1=data1'; data2=data2'; % transpose data to get it into a form acceptable to Murray's routine 0040 if nargin < 3, 0041 T1 = [min(data1(:,1)) max(max(data1))]; 0042 end 0043 if nargin < 4, 0044 T2 = T1; 0045 end 0046 if nargin < 5, 0047 parametric = 0; 0048 end 0049 if nargin < 6; p = 0.05;end 0050 if nargin < 7; quiet = 0; end 0051 0052 if isempty(T1), 0053 T1 = [min(data1(:,1)) max(max(data1))]; 0054 end 0055 if isempty(T2) 0056 T2 = T1; 0057 end 0058 if isempty(parametric), 0059 parametric = 0; 0060 end 0061 if isempty(p) 0062 p = 0.05; 0063 end 0064 if isempty(quiet), 0065 quiet = 0; 0066 end 0067 0068 NT1 = length(data1(:,1)); 0069 NT2 = length(data2(:,2)); 0070 0071 if (NT1 < 4 || NT2 < 4) && parametric ~= 2, 0072 disp('Low number of trials : switch to Poisson test') 0073 parametric = 2; 0074 end 0075 0076 if abs((T1(2)-T1(1)) - (T1(2)-T1(1))) > 10.^-6 0077 error('Time intervals for analysis are different') 0078 end 0079 0080 % get counts... 0081 N1=zeros(1,NT1); 0082 for n=1:NT1 0083 N1(n) = length(find(data1(n,:) >= T1(1) & ... 0084 data1(n,:) <= T1(2) & ~isnan(data1(n,:)))); 0085 end 0086 N2=zeros(1,NT2); 0087 for n=1:NT2 0088 N2(n) = length(find(data2(n,:) >= T2(1) & ... 0089 data2(n,:) <= T2(2) & ~isnan(data1(n,:)))); 0090 end 0091 M1 = mean(N1); 0092 M2 = mean(N2); 0093 0094 % do non parametric test... 0095 0096 if parametric == 0 0097 [P H] = ranksum(N1,N2,p); 0098 end 0099 0100 % parametric test (with stabilizing transform)... 0101 0102 % use sqrt transformation from 0103 % Cox and Lewis to make data more Gaussian 0104 % the statistical analysis of series of events pg 44 0105 0106 if parametric == 1 0107 X = sqrt(N1 +0.25); 0108 Y = sqrt(N2 +0.25); 0109 [H,P] = ttest2(X,Y,p,0); 0110 end 0111 0112 % Poisson test. Use method from Zar 0113 % pg 580 Ed 3. Z bahaves as a normal variate under 0114 % null of same process mean and Poisson processes 0115 0116 if parametric == 2 0117 X = sum(N1); 0118 Y = sum(N2); 0119 Z = abs(X-Y)./sqrt(X+Y); 0120 P = 2*(1-normcdf(Z)); 0121 if P < p; H = 1;else H = 0;end 0122 end 0123 0124 if quiet == 0 0125 if H == 1 0126 disp('Counts are signifcantly different') 0127 else 0128 disp('Counts are not signifcantly different') 0129 end 0130 disp(['Mean count for data1 = ' num2str(M1)]) 0131 disp(['Mean count for data2 = ' num2str(M2)]) 0132 disp(['P value = ' num2str(P)]) 0133 end 0134 0135 0136 0137 0138 0139 0140 0141 0142 0143 0144