Here's a start
% Construct time axis (called X instead of t for some reason).
X = -0.5 : 0.001 : 0.5
% Construct Y that is the same length as X.
Y = rand(size(X))
% Create filter window:
kernel = ......
% Do the convolution
yFiltered = conv(Y, kernel)
If you need more help, read this:
Another hint: create a step function with the edge at 100, then inverse fft it to get the spatial domain kernel that you convolve the signal with.