Hey Raja,
In the case your foundry does not provide you this data, you can extract it out by running a set of simulations. In the past, I ran >500 iterations to extract AVT for a device. Here is how the flow went,
If you have monte-carlo simulator (ADEXL or BDA), then you can create a testbench to extract AVT out. Make your "L" and "W" parameters for a standard NCH device, create a dcOp test bench. Create a measure to extract the threshold (VT) of the device for a given W/L. Once you have this, run a monte-carlo for sample size of 100 iterations. Observe the standard deviation (1-sigma) of your VTH measurements. Now, vary your W and or L and re-capture a new sigma.
If you tabulate enough points, you can try to fit the sigma_measure = AVT/sqrt(W*L) expression in Excel.
Hope this helps,
-Gujaratibhai