Question 1
26 marks
Cracked components: a chi-squared test, a binomial test and a two-sample t-test
- Statistics and probability
(a)(i)
1 mark
The total number of components with no cracks is 141, and 88 of them were made by the current technique.
(a)(ii)
1 mark
The 120 components made with the new technique are split between none, minor and major cracks.
(b)
2 marks
150 components had minor cracks and 54 of them were made with the new technique.
(c)(i)
1 mark
- : the level of cracking (none, minor, major) is independent of the production technique.
- : the level of cracking is not independent of the production technique.
(c)(ii)
2 marks
Enter the table into the calculator for a test (the smallest expected count is 23, so the test is valid). The degrees of freedom are .
(c)(iii)
2 marks
Since we reject . There is sufficient evidence that the amount of cracking depends on the production technique.
(d)
1 mark
For the current technique, of the 240 components developed cracks.
(e)
1 mark
The distribution assumes that the components are independent of each other (whether one cracks does not affect whether another does) and that each has the same probability of cracking.
(f)
5 marks
- The number of components with cracks among the 120 new ones is . Under , , and the alternative is that fewer crack.
- : , : .
- Since we do not reject . There is insufficient evidence at the 5% level that the new technique reduces the proportion of components that develop cracks.
(g)(i)
1 mark
The test in (f) is one-tailed and answers the question that matters here directly: whether the new technique reduces the number of components that crack. The test in (c) only says there is some difference.
(g)(ii)
1 mark
The test in (f) treats the proportion for the current technique (19/30) as an exact known value, but it is only an estimate from a sample, so could be wrong. It also lumps minor and major cracks together, losing information that the test in (c) uses.
(h)
7 marks
- Let and be the population mean times before cracking for the new and the current technique.
- Use a one-tailed two-sample -test with the sample sizes 100 and 200, means 250.1 and 249.1 and standard deviations 3.8 and 3.8 (pooled, since the standard deviations are equal).
- Since , reject . There is sufficient evidence that the new technique increases the mean time before cracks appear.
(i)
1 mark
Although statistically significant, the mean time has only gone up by about 1 hour out of about 250, a tiny effect. And a company would care about the smallest times before cracking (when parts need replacing) as well as the mean, so the statement overstates what the test shows.

