snanalyst/bootcamp

bootcamp

Companion package for the JADS Stats Bootcamp

bootcamp's Issues

Days 3 and 4 tutorials

1. Typo: First, in Day 3 | Reading Results, it states for the exponentiation of log of odds -2.55 that this makes it '0.78' more likely, rather than '0.078'.

2. The information for Day 4 | Statistically Significant with the alternative = "less" or "greater" is confusing, including the last paragraph corresponding to the visualization (red part) -- furthermore, the explanation belonging to the two independent t-tests with "greater" and "less" also seems contradictive (C < T is the same as T > C)

refer to this: a one-tailed t.test specified as alternative = "less" is testing if the mean of the control group is less than the mean of the treatment. As a consequence, a one-tailed t.test specified as alternative = "greater" is testing if the mean of the control group is greater than the mean of the treatment.

Also, the phrasing of the two sentences in the tutorial with less and greater end up saying in both cases that m_0 < m1. they should say two different cases instead.

3. Finally, in the interpretation of the Linear Regression (Day 4 | Linear Regression), it states that one group is 249 times more likely, rather than having 249 more clients (outcome variable), which I think should be correct.

add tests for the tutorials

like for the snatutti package, we need to have the tests included for the bootcamp tutorials

add unit tests for the functions

include unit tests for the functions included in the bootcamp package

bootcamp::desciptives, shapiro test applicability

The functions calls stats::shapiro.test which only accepts numeric vectors of sizes between 3-5000. When a longer data vector is passed, it outputs an error.
This is by design, as the calculation of the p-value of the statistic is not validated for sample sizes beyond 5000.
See: or .

The result is that the bootcamp::compareLM errors out when a vector larger than 5000 is passed.
There are a few solutions:

• replace the S&W test with an alternative that does work for larger samples, like nortest::ad.test
• prevent the error by tryCatching it first and then report NA in the output