Running README example requires extra dependency

[erikcs]

After I installed medoutcon from github, a big handful of dependencies, and dependencies' dependencies were updated/installed. When I run the readme example I get the following, which may seem a bit quirky for a new user:

> # compute targeted minimum loss estimate of the interventional direct effect
> tmle_de <- medoutcon(W = example_data[, ..w_names],
+                      A = example_data$A,
+                      Z = example_data$Z,
+                      M = example_data[, ..m_names],
+                      Y = example_data$Y,
+                      effect = "direct",
+                      estimator = "tmle")
Error in requirePackages(private$.required_packages, why = class(self)[1],  : 
  For Lrnr_glm_fast please install the following packages: speedglm

[nhejazi]

Thanks for catching this and pointing it out. This seems to happen because the default algorithms used for nuisance estimation wrap the speedglm package via sl3 (https://github.com/nhejazi/medoutcon/blob/master/R/medoutcon.R#L81-L85), while speedglm is only listed in the Suggests of this package. This could be fixed by moving speedglm to Depends, I believe. The same is true of hal9001 also (https://github.com/nhejazi/medoutcon/blob/master/R/medoutcon.R#L86-L87).

[nhejazi]

Following up on this to note that 4d5d15a in #34 moved both hal9001 and speedglm to the Depends slot. In testing the README example in a clean R session (R --vanilla), the example appears to work without issue; that is, I cannot reproduce the error. If you could spare a minute to try with that package version and confirm, that'd be helpful. If not, I'll try to run on another machine just to double check. Thanks again for your review of this package @erikcs.

[erikcs]

Hi @nhejazi, confirming this works here.

[nhejazi]

OK, I've now tried the example again on a completely fresh machine and believe that this issue has been resolved.

[nhejazi]

Great, thanks for confirming @erikcs!