Currently in boot_networklevel the seed is set internally and maybe that is not a good practice. Reproducibility is assured, but maybe should let the user be a bit more in charge of this.
In boot_networklevel there are two cases when seed is employed:
- In the usage of
boot_networklevel_once, when computing a network index for the sampled interactions. Is set internally to set.seed(42);
- Then in
boot_networklevel_n, when calling sample_indices;
- Again in
boot_networklevel_n, when running boot_networklevel_once in parallel an the iterator is passed as seed to boot_networklevel_once.
I think point 3 should stay unchanged because each iteration should generate different samples, otherwise we take the same sample every time. Or could generate a sequence of integers of the same length as the amount of iterators based on a user given seed and then pass those as seeds to boot_networklevel_once. However, that sounds unnecessarily complicated...
Nevertheless, the seed at point 2 could be a parameter that the user may want to change. Would this be worth implementing? If yes, then set the default value of 42 so that users don't snap if they test old results with new ones.
This should also be applicable for boot_specieslevel.
Currently in
boot_networklevelthe seed is set internally and maybe that is not a good practice. Reproducibility is assured, but maybe should let the user be a bit more in charge of this.In
boot_networklevelthere are two cases when seed is employed:boot_networklevel_once, when computing a network index for the sampled interactions. Is set internally toset.seed(42);boot_networklevel_n, when callingsample_indices;boot_networklevel_n, when runningboot_networklevel_oncein parallel an the iterator is passed as seed toboot_networklevel_once.I think point 3 should stay unchanged because each iteration should generate different samples, otherwise we take the same sample every time. Or could generate a sequence of integers of the same length as the amount of iterators based on a user given seed and then pass those as seeds to
boot_networklevel_once. However, that sounds unnecessarily complicated...Nevertheless, the seed at point 2 could be a parameter that the user may want to change. Would this be worth implementing? If yes, then set the default value of 42 so that users don't snap if they test old results with new ones.
This should also be applicable for
boot_specieslevel.