Defaults for all arguments in exported functions

[joshwlambert]

Would it be a (reasonable) option to add defaults for all arguments here?

Thinking about reducing barriers to usage, being able to run sim_linelist() and getting a line list returned (even if non-informative) may prove helpful to encourage usage. Now it requires going through additional steps that are not trivial - this can increase the barrier to first use enough to just move on if people are pressed for time.

Vice versa, the more accomplished we can make people feel in starting to use the package, the better from my perspective 😄

Originally posted by @chartgerink in #117 (comment)

[joshwlambert]

Comment from @chartgerink from PR #117 relevant to this discussion:

It would be great if getting started only requires

library(simulist)
sim_linelist()

This will encourage the next steps to specify the distributions. Right now this quick start feels not so quick and may be a barrier for people.

[joshwlambert]

I'm happy to implement this change. There is the risk that users will use the functions inappropriately with the default settings and the current setup provides some desirable difficulty, however, as this package is simulating data and not an analytical method used for estimation or inference then inappropriate use should be less problematic.

Discussion on this topic from Epiverse-TRACE (that I could find) are:

• Easy route to homogenous estimate finalsize#180
• Remove infection class epidemics#140

which seems to be in favour of providing sensible defaults over having users specify arguments. @Bisaloo do you agree that {simulist} would be worth adding more argument defaults to; and more generally is it Epiverse policy to use defaults wherever possible?

The arguments that would need defaults adding to achieve the simplest function call sim_linelist() outlined by @chartgerink would be:

For sim_linelist() and sim_outbreak():

• contact_distribution
• infectious_period
• prob_infection
• onset_to_hosp
• onset_to_death

For sim_contacts():

• contact_distribution
• infectious_period
• prob_infection

I propose to use the same anonymous functions as are used in the Get Started vignette.

Also tagging @CarmenTamayo and @adamkucharski to get other opinions from those involve in development discussions.

[Bisaloo]

@Bisaloo do you agree that {simulist} would be worth adding more argument defaults to; and more generally is it Epiverse policy to use defaults wherever possible?

My opinion is not fully formed on this and may continue evolving but my main intuition at this time, which results in this somewhat vague guidance:

Default are fine with the following rules and caveats:

• Scientifically valid defaults should be provided when possible (e.g., probably don't default to 100 iterations in a MCMC)
• If you can't provide scientifically valid defaults, avoid providing defaults at all for "difficult" parameters. I suspect some users start incorrectly relying on the default because they are not always sure what they mean or how to set them to a meaningful value.

[joshwlambert]

Thanks @Bisaloo. Following your suggestions I think the sim_*() functions in {simulist} are good candidates to have default arguments and I'll make this change.

[joshwlambert]

Closing as default arguments added in PR #149.