Help need for clarifying linear and nonlinear device model ?

Dear Sir,
I designed low noise amplifier at 1.3GHz using linear transistor model Avago ATF 54143 , and obtained Gain, Noise Figure and Stability. Now I want to check the linearity of the LNA using IIP3, OIP3, and P1 dB compression point. I have to use nonlinear device model ATF 54143 to measure the IIP3, OIP3, P1dB compression point. What is my doubt is I have connected the nonlinear device model to same circuit and biasing, whatever the circuit and biasing used for the linear device model. I am getting the entirely different results for both linear and nonlinear device model for same circuit and biasing. What is my question is do I need to use the nonlinear device model for measuring IIP3,OIP3, P1dB , Gain and NF? or do I need to use only for linear device model for measuring Gain , Noise Figure and stability, and do I need to use nonlinear device model for measuring only IIP3, OIP3, P1dB or do I need to use the nonlinear device model for measuring both IIP3 , OIP3,P1dB, Gain and Noise Figure and Stability. Can you anybody help me to sort out.

Both linear and non-linear simulation of a transistor should produces very similar results in terms of gain and stability. This is valid if the bias point (Vds,Id) in the non-linear circuit is the same of which used for s-parameter exctraction (linear model) supposing all the other parts of the two circuits are the same.

albbg Sir,
I checked both linear and nonlinear with same input and output matching and same biasing, am getting the entirely different results. I couldn't find , where I am making mistake. I set Vds =3.02 V and Id=63.7mA.

Both the simulation of the two images you post, seems to me linear.
However in a linear simulation where the active component is modeled by s-parameter, you cannot change the bias point. It is that under which the s-parameter have been extracted. This means the voltage source used to supply the s-parameter block will be substituted by a short-circuit.