Whatever statistical model we fit, we progress through a workflow that can be remembered by knowing that model fitting is E.V.I.L.:
[E]{.pkg}valuate the model
[V]{.pkg}isualise the data/hypotheses
[I]{.pkg}nterpret the model
[L]{.pkg}oad the data and [L]{.pkg}ook at descriptive summaries
We can can use a consistent set of steps and r rproj() functions as highlighted in Figure 2, although the order of steps is a bit more fluid than this list implies. For example, [V]{.alt}isualise is often done after [E]{.alt}valuating the model.
[Step 1]{.alt}. [L]{.pkg}oad the data into r rproj(), pre-process and summarize using [tidyverse]{.pkg} packages or the [datawizard]{.pkg} package from [easystats]{.pkg}.
[Step 2]{.alt}. [V]{.pkg}isualise the data using the [ggplot2]{.pkg} package from [tidyverse]{.pkg}.
[Step 3]{.alt}. Fit a statistical model that will answer your research question.
[Step 4]{.alt}. [E]{.pkg}valuate the model using a small number of functions from the [performance]{.pkg} package from [easystats]{.pkg}. Specially, we evaluate
How well the model fits the data
Whether the underlying assumptions of the model are met.
[Step 5]{.alt}. [I]{.pkg}nterpret the model
Interpret the values that define the model (the [parameter estimates]{.alt}) using the [parameters]{.pkg} package from [easystats]{.pkg}.
Make specific predictions from our model using the [modelbased]{.pkg} package from [easystats]{.pkg}.
Figure 2: Workflow for fitting statistical models.Figure 3: What to do if the model assumptions are broken.
