Statistical Process Control
1. The Wall Street Journal article (Sept. 24 2012) notes how “American Airlines continued to rack up high numbers of flight delays and cancellations, blaming a dispute with its pilots union. The union, meanwhile, denied that pilots disrupted flights unnecessarily."
"By my calculations," writes a Dallas Morning News (Sept. 20, 2012) reporter, "the number of pilots on sick leave was 11 pilots higher on Sept. 18, 2012, than on Sept. 18, 2011(1). That seems like an increase in sick leave usage". (See the bar chart graph below used to make this point).
Can the same set of data be used to make opposite points in an argument? It's not that statistics lie, it is more in how we present all of the available data points, as can see in this example regarding the alleged "sick out" of American Airlines pilots. Here is a 13 month "snapshot" of percent of pilots out sick at American Airlines:
Date % Sick
9/18/2011 5.0%
10/18/2011 9.5%
11/18/2011 5.0%
12/18/2011 6.4%
1/18/2012 5.4%
2/18/2012 6.6%
3/18/2012 6.6%
4/18/2012 5.9%
5/18/2012 7.1%
6/18/2012 7.3%
7/18/2012 6.1%
8/18/2012 6.6%
9/18/2012 7.6%
Counters the union: “Contrary to claims by management, we have confirmed that pilot sick rates have not deviated from historical norms".
a) Do you agree with the union if the lower-control limit (LCL) is 2.94%, and the upper-control limit (UCL) is 10.15% given the 3-sigma p-chart showing the percent of pilots sick? (Hint: See note (1) for the sample size). Do not forget to plot the center line.
b) What happens if you use 2-sigma control limits? Recalculate the control limits and discuss how your interpretation changes and why?
c) Alternatively, what happens if you exclude the first two months? Recalculate the control limits and discuss whether your interpretation changes. If so, how?
"By my calculations," writes a Dallas Morning News (Sept. 20, 2012) reporter, "the number of pilots on sick leave was 11 pilots higher on Sept. 18, 2012, than on Sept. 18, 2011(1). That seems like an increase in sick leave usage". (See the bar chart graph below used to make this point).
Can the same set of data be used to make opposite points in an argument? It's not that statistics lie, it is more in how we present all of the available data points, as can see in this example regarding the alleged "sick out" of American Airlines pilots. Here is a 13 month "snapshot" of percent of pilots out sick at American Airlines:
Date % Sick
9/18/2011 5.0%
10/18/2011 9.5%
11/18/2011 5.0%
12/18/2011 6.4%
1/18/2012 5.4%
2/18/2012 6.6%
3/18/2012 6.6%
4/18/2012 5.9%
5/18/2012 7.1%
6/18/2012 7.3%
7/18/2012 6.1%
8/18/2012 6.6%
9/18/2012 7.6%
Counters the union: “Contrary to claims by management, we have confirmed that pilot sick rates have not deviated from historical norms".
a) Do you agree with the union if the lower-control limit (LCL) is 2.94%, and the upper-control limit (UCL) is 10.15% given the 3-sigma p-chart showing the percent of pilots sick? (Hint: See note (1) for the sample size). Do not forget to plot the center line.
b) What happens if you use 2-sigma control limits? Recalculate the control limits and discuss how your interpretation changes and why?
c) Alternatively, what happens if you exclude the first two months? Recalculate the control limits and discuss whether your interpretation changes. If so, how?