```{r data generation, echo = FALSE, results = "hide"}
n <- sample(8:15, 1)
y <- rnorm(n, runif(1, 4940, 4990), runif(1, 30, 50))
alpha <- sample(c(0.1, 0.05, 0.01), 1)

Mean <- round(mean(y), digits = 1)
Var <- round(var(y), digits = 1)
sd <- sqrt(Var/n)
fact <- round(qt(1 - alpha/2, df = n - 1), digits = 4)
facn <- round(qnorm(1 - alpha/2), digits = 4)
LBt <- round(Mean - fact * sd, digits = 3)
UBt <- round(Mean + fact * sd, digits = 3)
LBn <- round(Mean - facn * sd, digits = 3)
UBn <- round(Mean + facn * sd, digits = 3)

## use extended Moodle processing to award 100% for correct solution based on
## t quantiles and 50% for the solution based on normal quantiles
##
## this can be handled as a "verbatim" solution, directly including Moodles
## cloze type:
##   ":NUMERICAL:=2.228:0.01~%50%1.960:0.01#Normal-based instead of t-based interval."
## where 2.228 is the correct and 1.960 the partially correct solution,
## the tolerance is 0.01 in both cases, and a comment is supplied at the end.
## More details: https://docs.moodle.org/35/en/Embedded_Answers_(Cloze)_question_type

## solution template (note: % have to be escaped as %% for sprintf)
sol <- ":NUMERICAL:=%s:0.1~%%50%%%s:0.1#Normal-based instead of t-based interval; for small samples, intervals based on the normal approximation are too narrow."

## insert correct and partially correct solutions
sol <- sprintf(sol, c(LBt, UBt), c(LBn, UBn))
```

Question
========
It is suspected that a supplier systematically underfills 5 l canisters of detergent.
The filled volumes are assumed to be normally distributed. A small sample of $`r n`$
canisters is measured exactly. This shows that the canisters contain on average
$`r Mean`$ ml. The sample variance $s^2_{n-1}$ is equal to $`r Var`$.

Determine a $`r 100 * (1 - alpha)`\%$ confidence interval for the average content of
a canister (in ml).

Answerlist
----------
* What is the lower confidence bound?
* What is the upper confidence bound?

Solution
========
The $`r 100 * (1 - alpha)`\%$ confidence interval for the average content $\mu$ in ml is
given by:
$$
\begin{aligned}
& \left[\bar{y} \, - \, t_{n-1;`r 1-alpha/2`}\sqrt{\frac{s_{n-1}^2}{n}}, \;
  \bar{y} \, + \, t_{n-1;`r 1-alpha/2`}\sqrt{\frac{s_{n-1}^2}{n}}\right] \\
&= \left[ `r Mean` \, - \, `r fact`\sqrt{\frac{`r Var`}{`r n`}}, \;
          `r Mean` \, + \, `r fact`\sqrt{\frac{`r Var`}{`r n`}}\right] \\
&= \left[`r LBt`, \, `r UBt`\right].
\end{aligned}
$$

Answerlist
----------
* The lower confidence bound is $`r LBt`$.
* The upper confidence bound is $`r UBt`$.


Meta-information
============
extype: cloze
exclozetype: verbatim|verbatim
exsolution: `r sol[1]`|`r sol[2]`
exname: Confidence interval
extol: 0.01