My paper is accepted with minor revisions and now is in proof stage. Unfortunately, I have found computational errors in my calculations which have led to some wrong results.
What should I do?
Sometimes mistakes happen. Yes, it's a pain to have to deal with this mistake now after the paper was accepted for publication, but it's a good thing that you caught your mistake before the paper was published.
Note: If you have coauthors, before doing anything, contact your coauthors and explain to them the changes you need to make. Do this before doing anything else.
As you mentioned that you are dealing with an IEEE journal, my response here is specific to the IEEE:
Talk to the senior editor that you are working with currently. Tell them specifically the changes you want to make and that "the logic of the concluding remarks is still correct and the discrepancies only impact the affected numerical results and the specific conclusions drawn from them," and that "the other numerical results are correct."
The senior editor will take it from there. Based on my past experience, you do not need to contact the editor-in-chief about this directly.
No, but I don't really follow why you think correcting a mistake will affect your paper's value. Correcting a mistake prior to publication is a good thing.
This will depend on the steps taken by the senior editor.
If your findings and conclusions are unambiguously unaffected, send the results to the editor saying you found an error with the numbers and asking for it be fixed. The editor might be annoyed at the last minute change but they are used to this sort of thing. I don't know what your "paper value" refers to.
You seem pretty worried about this. If this is because you think other people might assume that this difference would, in fact, have affected the findings and the way in which your paper was reviewed? If so, that is an issue. If this is a concern, you should email the academic editor and explain the situation. If things are as you think, there won't be an issue. If the editor disagrees, the paper might be need to be re-reviewed. In any case, it certainly seems best to have this all above board.
Changes in language, like changes in numbers, can have an important impact on a paper. As an author, you have a responsibility to not make substantively important changes to either after the paper has been accepted. It's your responsibility to not abuse the editorial system. If you are worried that you might be, you need to talk to your editor.
If you really care about the validity of your data and that it may be used for future research by yourself and others, I would say that it is vital to send the errata to the editor and you will feel better for three reasons:
It will probably have to be re-reviewed, but if you're up front and honest with the editor or copy-editor about the errata, it's unlikely they'll reject the paper.