easystats.bsky.social
Official channel of {easystats}, a collection of #rstats 📦s with a unifying and consistent framework for statistical modeling, visualization, and reporting
40 posts
746 followers
110 following
Regular Contributor
Conversation Starter
comment in response to
post
See also the many ways to sutomize the output:
easystats.github.io/parameters/a...
Or the many table creating packages, like modelsummary.com or www.danieldsjoberg.com/gtsummary/, if the customization options in {parameters} are not sufficient.
comment in response to
post
It's comparable to the {broom} package, but,
- `model_parameters()` prints nicely
- has many additional features, like standardizing or robust standard errors
- allows different approximation methods for degrees of freedom (Satterthwaite, ...)
and more! See also easystats.github.io/parameters/a...
comment in response to
post
HC standard errors are unfortunately not available for glmmTMB. Would it be possible to model a dispersion parameter in `glmmTMB()` instead? Anyway, thanks for the feedback, we added a short section on standard error estimators.
comment in response to
post
Thanks for pointing this out! At least for this particular example, bootstrapping returned a smaller confidence interval range (i.e. lower SEs). Not sure about the general behaviour for bootstrapped SEs. Maybe go Bayesian, because bootstrapping mixed models is really slow...
comment in response to
post
The `model_dashboard()` in action... #easystats #rstats
comment in response to
post
Try it out now! doi.org/10.32614/CRA...
comment in response to
post
Plotting gives you nice figures out of the box!
comment in response to
post
The output is clear and informative, see following examples:
comment in response to
post
The package is designed around three core ideas:
1) Understand your results with marginal means and marginal effects: `estimate_means()` and `estimate_slopes()`
2) Check for differences between groups in your samle: `estimate_contrasts()`
3) Communicate your results with plotting: `plot()`
comment in response to
post
We're happy that you like the packages and the concept! We also had the idea of a book, but unfortunately we don't have the time (or people who want to write it). We're starting with vignettes showing some workflows (easystats.github.io/easystats/ar...), maybe that could be compiled to a Quarto doc
comment in response to
post
Thanks, should be fixed in the current GitHub version of {parameters}.
comment in response to
post
You're late to the party, but not too late 😉
comment in response to
post
Found any inconsistencies across packages? Let us know!
comment in response to
post
comment in response to
post
That'd be great if some of our packages will be featured!
comment in response to
post
Or that for log-responses, the raw coefficient (a coefficient of -0.2 means ~20% decrease) is less accurate than exponentiating it and interpret the ratio (which is exp(-0.2) ~ 0.82 ~ 18% decrease).
comment in response to
post
Yes, interpretation could better be moved to the docs. Question is, since some statistical packages report both, say, log-odds and odds ratios, it might be helpful to tell users that there's an exponentiate-option.
comment in response to
post
Sounds reasonable!