constructstimulusmatrices

 function f = constructstimulusmatrices(m,prenumlag,postnumlag,wantwrap)

 <m> is a 2D matrix, each row of which is a stimulus sequence (i.e.
   a vector that is all zeros except for ones indicating the onset
   of a given stimulus (fractional values are also okay))
 <prenumlag> is the number of stimulus points in the past
 <postnumlag> is the number of stimulus points in the future
 <wantwrap> (optional) is whether to wrap around.  default: 0.

 return a stimulus matrix of dimensions
 size(m,2) x ((prenumlag+postnumlag+1)*size(m,1)).
 this is a horizontal concatenation of the stimulus
 matrix for the first stimulus sequence, the stimulus
 matrix for the second stimulus sequence, and so on.
 this function is useful for fitting finite impulse response (FIR) models.

 history:
 2013/05/12 - update doc to indicate fractional values are okay.

 example:
 imagesc(constructstimulusmatrices([0 1 0 0 0 0 0 0 0; 0 0 1 0 0 0 0 0 0],0,3));

0001 function f = constructstimulusmatrices(m,prenumlag,postnumlag,wantwrap)
0002 
0003 % function f = constructstimulusmatrices(m,prenumlag,postnumlag,wantwrap)
0004 %
0005 % <m> is a 2D matrix, each row of which is a stimulus sequence (i.e.
0006 %   a vector that is all zeros except for ones indicating the onset
0007 %   of a given stimulus (fractional values are also okay))
0008 % <prenumlag> is the number of stimulus points in the past
0009 % <postnumlag> is the number of stimulus points in the future
0010 % <wantwrap> (optional) is whether to wrap around.  default: 0.
0011 %
0012 % return a stimulus matrix of dimensions
0013 % size(m,2) x ((prenumlag+postnumlag+1)*size(m,1)).
0014 % this is a horizontal concatenation of the stimulus
0015 % matrix for the first stimulus sequence, the stimulus
0016 % matrix for the second stimulus sequence, and so on.
0017 % this function is useful for fitting finite impulse response (FIR) models.
0018 %
0019 % history:
0020 % 2013/05/12 - update doc to indicate fractional values are okay.
0021 %
0022 % example:
0023 % imagesc(constructstimulusmatrices([0 1 0 0 0 0 0 0 0; 0 0 1 0 0 0 0 0 0],0,3));
0024 
0025 % input
0026 if ~exist('wantwrap','var') || isempty(wantwrap)
0027   wantwrap = 0;
0028 end
0029 
0030 % get out early
0031 if prenumlag==0 && postnumlag==0
0032   f = m';
0033   return;
0034 end
0035 
0036 % do it
0037 num = prenumlag + postnumlag + 1;
0038 f = zeros([size(m,2) num*size(m,1)]);
0039 for p=1:size(m,1)
0040   f(:,(p-1)*num+(1:num)) = constructstimulusmatrix(m(p,:),prenumlag,postnumlag,wantwrap);
0041 end
0042 
0043 %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% HELPER FUNCTION
0044 
0045 function f = constructstimulusmatrix(v,prenumlag,postnumlag,wantwrap)
0046 
0047 % function f = constructstimulusmatrix(v,prenumlag,postnumlag,wantwrap)
0048 %
0049 % <v> is the stimulus sequence represented as a vector
0050 % <prenumlag> is the number of stimulus points in the past
0051 % <postnumlag> is the number of stimulus points in the future
0052 % <wantwrap> (optional) is whether to wrap around.  default: 0.
0053 %
0054 % return a stimulus matrix of dimensions
0055 % length(v) x (prenumlag+postnumlag+1)
0056 % where each column represents the stimulus at
0057 % a particular time lag.
0058 
0059 % input
0060 if ~exist('wantwrap','var') || isempty(wantwrap)
0061   wantwrap = 0;
0062 end
0063 
0064 % do it
0065 total = prenumlag + postnumlag + 1;
0066 f = zeros([length(v) total]);
0067 for p=1:total
0068   if wantwrap
0069     f(:,p) = circshift(v,[0 -prenumlag+(p-1)]).';
0070   else
0071     temp = -prenumlag+(p-1);
0072     if temp < 0
0073       f(1:end+temp,p) = v(1-temp:end);
0074     else
0075       f(temp+1:end,p) = v(1:end-temp);
0076     end
0077   end
0078 end