selective fading rayleigh
In channel model for wireless communication, noise is modeled separately to the fading. If you have flat fading channel plus noise, it means that you have to generate one random variable fulfilling rayleigh distribution and MULTIPLY the transmit signal with it, then you ADD the noise (usually Gaussian distributed).
If you have frequency selective fading channel, i.e. the channel is modeled as FIR filter with each tap fulfilling rayleigh distributed, then you have to CONVOLVE the transmit signal with this FIR filter, and then ADD the noise.
So, for either flat fading channel or frequency selective fading channel, the Gaussaian noise model is the same.
Rayleigh fading is only due to multipath signals and it is refered to as fast fading (small scale fading), whereas the distance between Tx & Rx introduce slow fading (large scale fading). You consider large scale fading for designing cell, power control, etc. But if you only consider designing equalizer or modulation technique, then you only consider small-scale fading (Rayleigh, Rice)
best