The front-end EMI filter will degrade the output impedance of the switching converter and can potentially bring instability when the loop is closed in certain conditions. To determine the open-loop output impedance with this filter, I recommend to go with the PWM switch model and gradually simplify the circuit with intermediate sanity check: make sure the plots you obtain from the simplified versions are always identical in phase in magnitude with the first one:
In this first picture, there is no EMI filter and the output impedance classically starts with the inductive ohmic loss \$r_L\$ paralleled with the load resistance. Then it is inductive, resonate and goes down again in the capacitive slope before ending up as the ESR of the capacitor \$r_C\$. From the large-signal model, use the small-signal version to start looking a potential analysis.
Now, you can add a front-end EMI filter and see the effects on both models:
Immediately you see the glitch in the phase and magnitude responses. This glitch is brought by the \$RLC\$ filter resonance which affects the source: without the filter, the node \$V_{in}\$ is at 0 V for the analysis and the current delivered by the buck as no effect on node a. With the \$RLC\$ front-end filter, this is no longer the case and now node a is modulated, affecting the whole transfer function. Damping is necessary and that is what I detail in my APEC 2017 seminar.
For the final schematic, rearrange and simplify the circuitry to put in a friendly shape where all is arranged in a meaningful. Make sure the final plot is rigorously identical to the first one and there you go:
You can install a test generator \$I_T\$ (the stimulus) producing a voltage response \$V_T\$ across the generator's terminals. The output impedance is \$Z_{out}(s)=\frac{V_T(s)}{I_T(s)}\$ You have to solve a 4th-order transfer function and I recommend you use the fast analytical circuits techniques or FACTs. You have an example you can download from my page.
