Public key collision: Question

Hi all,

I try to understand and got stuck with the following:

• If someone would try to produce a collision of a private key, this is almost impossible, correct?

• If someone would try to produce a collision of a public key, this, too, is almost impossible.

But if someone would be successful in creating a collision of a public key, how likely would it be that the private key would be different? Or, just for the maths of it: What are the mathematical odds that, if someone would find a collision of public keys, that the private key would be different from the other private key which was tied to the first public key?