


ftestc computes the F-statistic for sine wave in locally-white noise (continuous data).
[Fval,A,f,sig,sd] = ftest(data,tapers,Fs,fpass,pad,p)
Inputs:
data (data in [N,C] i.e. time x channels/trials) - required.
tapers (parameters for calculating tapers [NW,K]) - optional. Defaults to [3 5]
Fs (sampling frequency) -- optional. Defaults to 1.
fpass (frequency band to be used in the calculation in the form
[fmin fmax])- optional.
Default all frequencies between 0 and Fs/2
pad (padding factor for the FFT) - optional. Defaults to 0.
e.g. For N = 500, if PAD = 0, we pad the FFT
to 512 points; if PAD = 2, we pad the FFT
to 2048 points, etc.
p (P-value to calculate error bars for) - optional. Defaults to 0.05 (95% confidence).
Outputs:
Fval (F-statistic in frequency x channels/trials form)
A (Line amplitude for X in frequency x channels/trials form)
f (frequencies of evaluation)
sig (F distribution (1-p)% confidence level)
sd (standard deviation of the amplitude C) - optional

0001 function [Fval,A,f,sig,sd] = ftestc(data,tapers,Fs,fpass,pad,p) 0002 % ftestc computes the F-statistic for sine wave in locally-white noise (continuous data). 0003 % 0004 % [Fval,A,f,sig,sd] = ftest(data,tapers,Fs,fpass,pad,p) 0005 % 0006 % Inputs: 0007 % data (data in [N,C] i.e. time x channels/trials) - required. 0008 % tapers (parameters for calculating tapers [NW,K]) - optional. Defaults to [3 5] 0009 % Fs (sampling frequency) -- optional. Defaults to 1. 0010 % fpass (frequency band to be used in the calculation in the form 0011 % [fmin fmax])- optional. 0012 % Default all frequencies between 0 and Fs/2 0013 % pad (padding factor for the FFT) - optional. Defaults to 0. 0014 % e.g. For N = 500, if PAD = 0, we pad the FFT 0015 % to 512 points; if PAD = 2, we pad the FFT 0016 % to 2048 points, etc. 0017 % p (P-value to calculate error bars for) - optional. Defaults to 0.05 (95% confidence). 0018 % 0019 % 0020 % Outputs: 0021 % Fval (F-statistic in frequency x channels/trials form) 0022 % A (Line amplitude for X in frequency x channels/trials form) 0023 % f (frequencies of evaluation) 0024 % sig (F distribution (1-p)% confidence level) 0025 % sd (standard deviation of the amplitude C) - optional 0026 if nargin < 1; error('Need data'); end; 0027 [N,C]=size(data); 0028 if nargin<2; tapers=[3 5];end; 0029 if nargin<3; Fs=1; end; 0030 if nargin<4;fpass=[0 Fs/2];end; 0031 if nargin<5;pad=0;end; 0032 if nargin<6;p=0.05;end; 0033 tapers=dpsschk(tapers,N); % calculate the tapers 0034 [N,K]=size(tapers); 0035 nfft=2^(nextpow2(N)+pad);% number of points in fft 0036 [f,findx]=getfgrid(Fs,nfft,fpass);% frequency grid to be returned 0037 errorchk = 0; % set error checking to default (no errors calculated) 0038 if nargout <= 3 % if called with 4 output arguments, activate error checking 0039 errorchk = 0; 0040 else 0041 errorchk = 1; 0042 end 0043 Kodd=1:2:K; 0044 Keven=2:2:K; 0045 J=mtfftc(data,tapers,nfft);% tapered fft of data - f x K x C 0046 Jp=J(findx,Kodd,:); % drop the even ffts and restrict fft to specified frequency grid - f x K x C 0047 tapers=tapers(:,:,ones(1,C)); % add channel indices to the tapers - t x K x C 0048 H0 = squeeze(sum(tapers(:,Kodd,:),1)); % calculate sum of tapers for even prolates - K x C 0049 if C==1;H0=H0';end; 0050 Nf=length(findx);% number of frequencies 0051 H0 = H0(:,:,ones(1,Nf)); % add frequency indices to H0 - K x C x f 0052 H0=permute(H0,[3 1 2]); % permute H0 to get dimensions to match those of Jp - f x K x C 0053 H0sq=sum(H0.*H0,2);% sum of squares of H0^2 across taper indices - f x C 0054 JpH0=sum(Jp.*squeeze(H0),2);% sum of the product of Jp and H0 across taper indices - f x C 0055 A=squeeze(JpH0./H0sq); % amplitudes for all frequencies and channels 0056 Kp=size(Jp,2); % number of even prolates 0057 Ap=A(:,:,ones(1,Kp)); % add the taper index to C 0058 Ap=permute(Ap,[1 3 2]); % permute indices to match those of H0 0059 Jhat=Ap.*H0; % fitted value for the fft 0060 0061 num=(K-1).*(abs(A).^2).*squeeze(H0sq);%numerator for F-statistic 0062 den=squeeze(sum(abs(Jp-Jhat).^2,2)+sum(abs(J(findx,Keven,:)).^2,2));% denominator for F-statistic 0063 Fval=num./den; % F-statisitic 0064 sig=finv(1-p,2,2*K-2); % F-distribution based 1-p% point 0065 var=den./(K*squeeze(H0sq)); % variance of amplitude 0066 sd=sqrt(var);% standard deviation of amplitude 0067 if nargout==0 0068 plot(f,Fval); line(get(gca,'xlim'),[sig sig],'Color','r');xlabel('frequency Hz'); 0069 ylabel('F ratio'); 0070 end