Session 2 Exercises

In the publishr_workshop RProject, open the exercises/00-data_cleaning.qmd Quarto script. Before you make any changes, render it.

Data and Code

Add appropriate headers below to keep your work structured.

Exercise 1: A simple code chunk

Add a code chunk below that computes 2 + 2 and renders it. Make sure the chunk is both printed (code visible) and evaluated (output visible).

Exercise 2: Echo without eval

Change the chunk options on your Exercise 1 chunk so that the code is printed but not evaluated. Render and observe the difference.

Tip

Relevant chunk option: eval: false

Exercise 3: Silent chunk

Add a new code chunk that computes 10 * 10. Set the chunk options so that neither the code nor the output appears in the rendered document. Render and verify it is invisible.

Tip

Relevant chunk option: include: false

Exercise 4: Load data

Add a code chunk that loads the sample dataset. Set the chunk options so that the code is not printed in the rendered output but the chunk is evaluated.

```{r}
df_raw <- readr::read_csv(here::here("data", "raw", "data_example.csv"))
```

Tip

Relevant chunk option: echo: false

Exercise 5: Anonymise data

Add a code chunk that anonymises the px value.

```{r}
set.seed(416) # for reproducibility

px_lookup <- df_raw |>
  distinct(px) |>                             # one row per participant
  mutate(px = sample(1:n(), n(), replace = FALSE)) # randomly assign 1:N

df_clean <- df_raw |>
  left_join(px_lookup, by = "px") |>
  select(-px) |>                              # remove original identifier
  relocate(px) |>
  arrange(px)
```

Exercise 6: Save processed data

Save this anonymised data in data/processed.

```{r}
write_csv(df_clean, here::here("data", "processed", "data_clean.csv"))
```

Assign code chunks accordingly!!! Do you want to run all these steps each time you render, or is rendering a way to document a one-time process?

Exercise 7: Summary statistics

Load in your saved data (what should eval be?).

Create a code chunk that filters to the verb region and computes the mean and SD of first-pass reading time (gaze), storing them as named objects. Set the chunk so it runs silently.

```{r}
m_gaze  <- # YOUR CODE HERE
sd_gaze <- # YOUR CODE HERE
```

Tip

You will need filter(region == "verb"), mean(..., na.rm = TRUE), sd(..., na.rm = TRUE), and round(..., 1).

Exercise 8: Inline reporting

Report your values from Exercise 7 inline by completing the sentence below:

Mean first-pass reading time at the verb region was ___ ms (SD = ___ ms).

Exercise 9: Global eval

Add the following to your YAML execute: block:

execute:
  eval: false

Render. What happens? Which exercises break and why?

Now fix the problem by adding #| eval: true to the chunks that need to run (Exercises 4 and 5). Re-render and confirm everything works.

Note

This exercise illustrates why global chunk options need to be set carefully, and why it is often better to set eval: false only on specific chunks rather than globally.

Models

Exercise 1: Analysis script

Open the script analysis.qmd and run each code chunk manually. Then render it.

Exercise 2: Set up your manuscript script

Open manuscript.qmd. The global chunk option to hide code is already set. Render it to confirm it works.

Add inline code to fill in the following sentence (lines 137-139 in the Quarto script):

There was a main effect of lifetime, with longer first-pass reading times 
for dead versus living referents ($\beta$ = , SE = , 
*t* = , *p* ).

Tip

You will need broom.mixed::tidy() with effects = "fixed" and conf.int = TRUE, then filter(term == "tense1") and pull() for each value.

Exercise 3: Reporting other effects

So far our manuscript only reports effects of lifetime. Do the same for the Intercept and the effects of tense1 and trial_c. How you structure the sentences and paragraph is up to you.

Exercise 4: Fit a total reading time model

Switch back to analysis.qmd.

Filter df_clean to the verb region and apply contrast coding and trial centring as covered in the chapter, but this time use tt (total reading time) as your dependent variable instead of gaze. Exclude zeros as before.
Set sum contrast coding for lifetime and tense. Center trial.
Fit the following model:

```{r}
lmer_tt <- lmer(
  log(tt) ~ lifetime * tense + trial_c +
    (1 + lifetime | px) +
    (1 + tense | item),
  data    = df_verb,
  control = lmerControl(optimizer = "bobyqa",
                        optCtrl   = list(maxfun = 1e5))
)
```

Check for convergence warnings. Is the model singular?
Save the model:

```{r}
saveRDS(lmer_tt, here("output", "models", "lmer_tt.rds"))
```

Exercise 5: Report the total reading time model in your manuscript

Switch to manuscript.qmd.

Load the new model:

```{r}
lmer_tt <- readRDS(here("output", "models", "lmer_tt.rds"))
```

Extract the values needed for a methods paragraph:

```{r}
n_obs_tt       <- nobs(lmer_tt)
n_px_tt        <- summary(lmer_tt)$ngrps[["px"]]
n_items_tt     <- summary(lmer_tt)$ngrps[["item"]]
formula_tt     <- deparse1(formula(lmer_tt))
optimizer_tt   <- lmer_tt@optinfo$optimizer
n_total_tt     <- n_px_tt * n_items_tt
n_dropped_tt   <- n_total_tt - n_obs_tt
pct_dropped_tt <- round(n_dropped_tt / n_total_tt * 100, 1)
```

Use these objects to fill in the blanks in manuscript.qmd under the Total reading times section.
Extract and report the lifetime main effect inline, following the same pattern as Exercise 2.

References

--- editor: source execute: eval: false echo: fenced --- # Session 2 Exercises {.unnumbered} In the `publishr_workshop` RProject, open the `exercises/00-data_cleaning.qmd` Quarto script. Before you make any changes, render it. ## Data and Code *Add appropriate headers below to keep your work structured.* ### Exercise 1: A simple code chunk Add a code chunk below that computes `2 + 2` and renders it. Make sure the chunk is both printed (code visible) and evaluated (output visible). ### Exercise 2: Echo without eval Change the chunk options on your Exercise 1 chunk so that the code is printed but not evaluated. Render and observe the difference. ::: callout-tip Relevant chunk option: `eval: false` ::: ### Exercise 3: Silent chunk Add a new code chunk that computes `10 * 10`. Set the chunk options so that neither the code nor the output appears in the rendered document. Render and verify it is invisible. ::: callout-tip Relevant chunk option: `include: false` ::: ### Exercise 4: Load data Add a code chunk that loads the sample dataset. Set the chunk options so that the code is not printed in the rendered output but the chunk is evaluated. ```{r} df_raw <- readr::read_csv(here::here("data", "raw", "data_example.csv")) ``` ::: callout-tip Relevant chunk option: `echo: false` ::: ### Exercise 5: Anonymise data Add a code chunk that anonymises the `px` value. ```{r} set.seed(416) # for reproducibility px_lookup <- df_raw |> distinct(px) |> # one row per participant mutate(px = sample(1:n(), n(), replace = FALSE)) # randomly assign 1:N df_clean <- df_raw |> left_join(px_lookup, by = "px") |> select(-px) |> # remove original identifier relocate(px) |> arrange(px) ``` ### Exercise 6: Save processed data Save this anonymised data in `data/processed`. ```{r} write_csv(df_clean, here::here("data", "processed", "data_clean.csv")) ``` **Assign code chunks accordingly!!!** Do you want to run all these steps each time you render, or is rendering a way to document a one-time process? ### Exercise 7: Summary statistics Load in your saved data (what should `eval` be?). Create a code chunk that filters to the verb region and computes the mean and SD of first-pass reading time (`gaze`), storing them as named objects. Set the chunk so it runs silently. ```{r} m_gaze <- # YOUR CODE HERE sd_gaze <- # YOUR CODE HERE ``` ::: callout-tip You will need `filter(region == "verb")`, `mean(..., na.rm = TRUE)`, `sd(..., na.rm = TRUE)`, and `round(..., 1)`. ::: ### Exercise 8: Inline reporting Report your values from Exercise 7 inline by completing the sentence below: ``` Mean first-pass reading time at the verb region was ___ ms (SD = ___ ms). ``` ### Exercise 9: Global eval Add the following to your YAML `execute:` block: ```yaml execute: eval: false ``` Render. What happens? Which exercises break and why? Now fix the problem by adding `#| eval: true` to the chunks that need to run (Exercises 4 and 5). Re-render and confirm everything works. ::: callout-note This exercise illustrates why global chunk options need to be set carefully, and why it is often better to set `eval: false` only on specific chunks rather than globally. ::: --- ## Models ### Exercise 1: Analysis script Open the script `analysis.qmd` and run each code chunk manually. Then render it. ### Exercise 2: Set up your manuscript script Open `manuscript.qmd`. The global chunk option to hide code is already set. Render it to confirm it works. Add inline code to fill in the following sentence (lines 137-139 in the Quarto script): ``` There was a main effect of lifetime, with longer first-pass reading times for dead versus living referents ($\beta$ = , SE = , *t* = , *p* ). ``` ::: callout-tip You will need `broom.mixed::tidy()` with `effects = "fixed"` and `conf.int = TRUE`, then `filter(term == "tense1")` and `pull()` for each value. ::: ### Exercise 3: Reporting other effects So far our manuscript only reports effects of lifetime. Do the same for the Intercept and the effects of `tense1` and `trial_c`. How you structure the sentences and paragraph is up to you. --- ### Exercise 4: Fit a total reading time model Switch back to `analysis.qmd`. 1. Filter `df_clean` to the verb region and apply contrast coding and trial centring as covered in the chapter, but this time use `tt` (total reading time) as your dependent variable instead of `gaze`. Exclude zeros as before. 2. Set sum contrast coding for `lifetime` and `tense`. Center `trial`. 2. Fit the following model: ```{r} lmer_tt <- lmer( log(tt) ~ lifetime * tense + trial_c + (1 + lifetime | px) + (1 + tense | item), data = df_verb, control = lmerControl(optimizer = "bobyqa", optCtrl = list(maxfun = 1e5)) ) ``` 3. Check for convergence warnings. Is the model singular? 4. Save the model: ```{r} saveRDS(lmer_tt, here("output", "models", "lmer_tt.rds")) ``` --- ### Exercise 5: Report the total reading time model in your manuscript Switch to `manuscript.qmd`. 1. Load the new model: ```{r} lmer_tt <- readRDS(here("output", "models", "lmer_tt.rds")) ``` 2. Extract the values needed for a methods paragraph: ```{r} n_obs_tt <- nobs(lmer_tt) n_px_tt <- summary(lmer_tt)$ngrps[["px"]] n_items_tt <- summary(lmer_tt)$ngrps[["item"]] formula_tt <- deparse1(formula(lmer_tt)) optimizer_tt <- lmer_tt@optinfo$optimizer n_total_tt <- n_px_tt * n_items_tt n_dropped_tt <- n_total_tt - n_obs_tt pct_dropped_tt <- round(n_dropped_tt / n_total_tt * 100, 1) ``` 3. Use these objects to fill in the blanks in `manuscript.qmd` under the Total reading times section. 4. Extract and report the lifetime main effect inline, following the same pattern as Exercise 2. ## References {.unnumbered} ::: {#refs} :::