# Specifying qubits to achieve measurement error mitigation on Qiskit

I'm learning how to do error mitigation on Qiskit as my experiment result differs from the simulated result. I read the tutorial here, but I have some questions about it. If I have understood it correctly, we take each basis states and measure them, to find the mitigation matrix, then execute my experiment, obtain the result, then recover the correct result using the mitigation matrix. That means I have to execute two experiment. However, how do I ensure that the qubits that I use to find the mitigation matrix is the same as the qubits that I do my actual experiment?

For example, let's say I want to execute an experiment with 3 qubits. I first run a code like the one in the tutorial. The compiler (or whatever compiles my code into instructions) uses qubit A, B and C. Then, when I execute my actual experiment, the compiler may use qubit D, E and F. As the qubits are different, then I think the mitigation will not be valid? Am I correct? if so, then how do we solve this? Obviously we have to use the same backend but I think that is achievable, but I found no way to specify which qubits to use.

job = execute(my_circuit, seed_transpiler=123)

Alternatively, if you would like to specify the layout yourself, you can do this by providing an initial_layout to the transpiler, and then setting the optimization_level=0. There are many potential formats for the initial layout, but the easiest to understand is a dictionary of the qubits in your circuit, to the physical qubits you would like them to take, for example {qr[0] : 3, qr[1]: 2}.
my_layout = {qr[0] : 3, qr[1]: 2}