2. 95% confidence interval for difference in means = (-13.69,26.53)
zero is in this interval, so no evidence to reject the hypothesis
that the means of the two distributions are different at the 5%
level of significance (p-value = 0.497 which is not small).
Hence, on the basis of the data, we do not reject the hypothesis
that the treatment has zero effect.
Small samples, so assumes underlying distributions of breaking strengths
are approximately normal - seems safe as probability plots look OK.
3. 95% confidence interval for mean difference = (0.20,19.46)
zero is not in this interval, so we reject the hypothesis
that the mean difference is zero at the 5% level of significance
(p-value = 0.047 which is just less than 5%).
Hence, on the basis of the data, we do reject the hypothesis
that the treatment has zero effect and conclude that the treatment
has an effect.
Paired experiments are often more informative, as variability
between units (in this case strips) is eliminated by having the
paired measurements made on the same set of units.
Hence more reliable results can be obtained using smaller samples.
4. Estimate of intercept, alpha = -0.650 and estimate of slope, beta = 0.921.