In ihrke/astatur: Package for the book "Applied Statistics using R"

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Even though we were trying to get everything right the first time, small errors can (and do!) slip through the cracks. Here we keep an updated list of errors and mistakes that have been gracefully pointed out to us by attentive readers.

Problems in the text

Eq. 7.16: There is an error in Eq. 7.16: there should not be a subscript i on $\bar{Y}$. The error is in both the numerator and the denominator.
Figure 7.3:
This figure is, strictly speaking, not correct, and can be misleading. The distances shown are not TSS, ESS and RSS, but terms that - when squared and summed - contribute to these.
We have both found it useful for explaining the concept to our students. However, the current caption: the current “Visual representation of $R^2$” may be a little misleading and should read “Conceptual illustration of $R^2$” to emphasize the fact that it is a simplification for illustrative purposes.
In section 7.2.2 we write that “the square root of R is equivalent to the Pearson correlation between Y and X”. This is only true if r is positive. It would be better to say that $r^2 = R^2$, or that the square of r is equal to $R^2$.
Last sentence in section 7.2.2: $F$ should not be enclosed in parentheses.
Section 8.1, after Eq. 8.2 we write: The same interpretation applies to the coefficients ($\beta_2, \beta_3, ..., \beta_k$). The last beta in the parenthesis should have subscript K, not k.
Figure 8.1: There should not be a hat on the epsilon in the figure.
Section 12.4, after output of anova(mod.linear,mod.quadratic, mod.cubic) we write: "[...] but that the cubic model does not outperform the quadratic model, $\chi^2(1) = 0.27, p = 0.27$."
The Chi-square value is wrong. It should be 1.21, not 0.27 (which is the p-value, printed twice).
On page 367 we mention the function smc() that calculates the squared multiple correlation and we write that it is part of the dplyr package. This is incorrect, as smc() is implemented in the psych package.

Problems with code

Section 10.2, towards the end there is a code segment that reads:

library(dplyr)
library(astatur)
workout <- mutate(workout, healthage = health * age)

The mutate function may give an error (with recent versions of R and the dplyr package) that it cannot compute the product of health and age, since health is of type haven_labelled. The problem can be solved by applying the function zap_formats():

workout <- haven::zap_formats(workout)

On page 89, we write

country <- factor(c("Dutch", "Wales", "Scotland", "Holland",
                    "Netherlands", "United Kingdom", "England",
                    "Northern Ireland"))
country <- fct_collapse(country,
                        Netherlands = c("Friesland", "Holland", "Netherlands"),
                        `United Kingdom` = c("Wales", "Scotland",
                                             "United Kingdom", "England",
                                             "Northern Ireland"))

Running this code will collapse the factor to categories "Netherlands", "United Kingdom" and "Dutch". Our intention was that category "Dutch" from the factor specification should be subsumed in the "Netherlands" category. Therefore, "Dutch" should be relabeled as "Friesland" or "Dutch" should be included in the list in the second statement using fct_collapse().

In section 12.1.1 on page 327, we give a more complex example of using pivot_longer() to convert a wide data.frame into long format. The example is:

depression.wide %>% 
  pivot_longer(cols=starts_with("week") | starts_with("BDI"),
               names_to = c(".value", "session"),
               names_pattern="(BDI|week)([0-9+])",
               values_drop_na=T) %>%
  head()

There is a mistake in the line names_pattern="(BDI|week)([0-9+])". The regular expression [0-9+] only captures a single integer and hence the code will not work as intended as values such as "BDI10" or "week12" will be coded as session=1 instead of session=10 and session=12, respectively. The correct regular expression should be "(BDI|week)([0-9]+)" such that the full code-example should read

depression.wide %>% 
  pivot_longer(cols=starts_with("week") | starts_with("BDI"),
               names_to = c(".value", "session"),
               names_pattern="(BDI|week)([0-9]+)",
               values_drop_na=T) %>%
  head()