Study. A software company asks for volunteers among its employees and gets 90. It randomly assigns 45 of them to a four-day work week and 45 to the usual five-day week for eight weeks. A focus app records each person's average minutes of focused work per day (focus). The analyst compares the schedules (schedule: four or five).
Interpret the 95% confidence interval in the output in context. Be explicit about the order of subtraction.
> t.test(focus ~ schedule, data = trial)
Welch Two Sample t-test
data: focus by schedule
t = 2.2886, df = 87.584, p-value = 0.02451
alternative hypothesis: true difference in means between group four and group five is not equal to 0
95 percent confidence interval:
1.842208 26.157792
sample estimates:
mean in group four mean in group five
212 198
# group sizes: four 45, five 45; sample SDs: four 30, five 28
# estimated effect size d-hat = 0.48
Math shortcuts are on for STATS 250. Type ^ then digits for powers, _ then digits for subscripts.
2
Part II · Lec 08 · 6 pts
Write a complete conclusion for the test that also reports and interprets the estimated effect size.
> t.test(focus ~ schedule, data = trial)
Welch Two Sample t-test
data: focus by schedule
t = 2.2886, df = 87.584, p-value = 0.02451
alternative hypothesis: true difference in means between group four and group five is not equal to 0
95 percent confidence interval:
1.842208 26.157792
sample estimates:
mean in group four mean in group five
212 198
# group sizes: four 45, five 45; sample SDs: four 30, five 28
# estimated effect size d-hat = 0.48
Math shortcuts are on for STATS 250. Type ^ then digits for powers, _ then digits for subscripts.
3
Part II · Lec 14 · 6 pts
The CEO writes: "This proves the four-day week causes more focused work, and it will do the same for all 2,000 of our employees." Evaluate both halves of this statement using the study design.
> t.test(focus ~ schedule, data = trial)
Welch Two Sample t-test
data: focus by schedule
t = 2.2886, df = 87.584, p-value = 0.02451
alternative hypothesis: true difference in means between group four and group five is not equal to 0
95 percent confidence interval:
1.842208 26.157792
sample estimates:
mean in group four mean in group five
212 198
# group sizes: four 45, five 45; sample SDs: four 30, five 28
# estimated effect size d-hat = 0.48
Math shortcuts are on for STATS 250. Type ^ then digits for powers, _ then digits for subscripts.
4
Part II · Lec 11 · 6 pts
A manager says: "If we had made a 99% interval instead, it would include 0, so at the 99% level the four-day week has no effect." Is the first part right? Use the p-value to decide without new output, then explain what is wrong with the conclusion.
> t.test(focus ~ schedule, data = trial)
Welch Two Sample t-test
data: focus by schedule
t = 2.2886, df = 87.584, p-value = 0.02451
alternative hypothesis: true difference in means between group four and group five is not equal to 0
95 percent confidence interval:
1.842208 26.157792
sample estimates:
mean in group four mean in group five
212 198
# group sizes: four 45, five 45; sample SDs: four 30, five 28
# estimated effect size d-hat = 0.48
Math shortcuts are on for STATS 250. Type ^ then digits for powers, _ then digits for subscripts.
5
Part II · Lec 09 · 6 pts
The company plans a replication with only 20 volunteers per schedule and expects the same true difference and the same spread. Compared with the original study, how would (i) the power of the test and (ii) the width of the 95% interval likely change? Explain each.
> t.test(focus ~ schedule, data = trial)
Welch Two Sample t-test
data: focus by schedule
t = 2.2886, df = 87.584, p-value = 0.02451
alternative hypothesis: true difference in means between group four and group five is not equal to 0
95 percent confidence interval:
1.842208 26.157792
sample estimates:
mean in group four mean in group five
212 198
# group sizes: four 45, five 45; sample SDs: four 30, five 28
# estimated effect size d-hat = 0.48
Math shortcuts are on for STATS 250. Type ^ then digits for powers, _ then digits for subscripts.
One composed randomized experiment with new Welch t-test output and five written prompts graded by rubric elements: interval, conclusion with effect size, scope of inference, the interval-test link, and power. Real Part II format: 40 minutes, about 10 to read and 30 to write. For confounder-table practice, use the Lec 13 section and the Case Study 2 lab. Not an official exam. Original questions written in the style of this course. Not official exam content.