4
$\begingroup$

How can someone properly decide on an appropriate level of theory MM, QM... What are the checklists throughout the experiment (in-silico)? How to decide a proper control case/group? For example from personal experience, it happened to me to work without a control (ground truth). I have realized this makes it difficult for the results to stand. There is a general workflow bellow:
Question --> Sample --> Method --> Results
So sample should include the test case(s) and a control. Other principles may be method reproducibility. I am just realizing how general this can be and can extend to any scientific method in general or depending on the type of question. But I am interested in an in-silico view.

$\endgroup$
5
  • 1
    $\begingroup$ Do you want to design an experiment or a theoretical modeling simulation? $\endgroup$
    – Camps
    Jul 3 '20 at 21:06
  • 5
    $\begingroup$ The title suggests it's an in silico experiment (i.e. a numerical experiment). BND: Can you please tell us more about what precisely you want to do? I gave you a +1 and didn't issue a close vote, but I think you really have to give us more than just this vague bit you've given so far. $\endgroup$ Jul 4 '20 at 2:37
  • 3
    $\begingroup$ I agree with Nike. This is a very broad question especially without a specific project in mind. Arguably, a lot of what you are asking about can only really be acquired through experience/trial and error. $\endgroup$
    – Tyberius
    Jul 4 '20 at 5:13
  • 1
    $\begingroup$ @camps a theoretical modelling simulation. $\endgroup$
    – BND
    Jul 4 '20 at 13:54
  • 2
    $\begingroup$ @BND I appreciate that you tried to add more details, but you still haven't given a specific problem that you want to solve using in-silico experiments. For example you could ask "How can I properly design an experiment to test that one basis set family converges faster to the complete basis set limit, than a different one?" My answer might be to: (1) come up with a test set, (2) calculate or obtain from elsewhere, an accurate set of benchmark CBS results, (3) try basis set A and basis set B, and so on. You could do the same for DFT functionals. But right now you have not told us what you want $\endgroup$ Jul 5 '20 at 14:25