Space Shuttle O-ring Failure (continued) - CAstatistics

Statistical Thinking
Space Shuttle O-ring Failure
The night before the disaster, it was predicted that it would be a cold night
(30°F). Some NASA engineers advised against a launch because they thought
that O-ring failures were related to temperature. They studied data on O-ring
failures the night before and no discernable relationship was found between Oring failure and temperature.
Hogg-Ledoleter (1992), Applied Statistics for Engineers and Physical Scientist.
No discernable pattern between number of O-ring failures and temperature is
present. Based on this information it was decided to launch.
Space Shuttle O-ring Failure (continued)
What went wrong? Quite simply, the engineers did not look at all the
available data. If they had examined O-ring successes and failures the night
before, they would have seen this:
Hogg-Ledoleter (1992), Applied Statistics for Engineers and Physical Scientist.
The extra data added is of flights with no incidents or O-ring failures. While O-ring
failures did occur at different temperatures, there is a distinct increased probably of
failure at lower temperature, and note that 30°F is not even on this scale.
Old Faithful Geyser
Centers and Averages
Prison Data
We have a sample of 30 countries reporting the number of people in prison
per 100,000 of the population. (JMP Datafile)
The data is a SAMPLE from 2007 collected from King’s College, London.
Questions to Students
1. Which is the better measure of central tendency, mean or median?
2. What are the characteristic(s) of countries with high or low prison
population rates?
3. Would you rather have the country you live in to have a high or low
prison population rate? Justify your answer.
Natural Variability
When does data indicate ‘Natural Variability’ or a distinct trend?
Exercise - 'Dodgy Dice'
We know events such as rolling dice and flipping coins have outcomes within the rules of natural variability.
Dice and coins have been used for gambling for thousands of years. Once in a while, a 'loaded' or 'dodgy' coin
or dice is used. This item is biased such that it does not have an expected value that a 'fair' item would. For
example, a biased coin might have the proportion of expected heads to be greater than 0.5 or a die might have
the expected roll of getting a '6' to be greater than 1 in 6. This gives someone an unfair advantage. The dataset
below the results from 50 rolls or flips of coins or dice.
What you need to determine is if the results indicate 'natural variation' or something 'dodgy'.
Column 1: 50 flips of a coin. A head is indicated with a '1', a tail with a '0'.
Column 2: 50 rolls of a die. The score is recorded.
Column 3: 50 rolls of two die. The total score is recorded.
Column 4: 50 rolls of a die. A '6' is indicated with a '1', a everything else with a '0'.
How do you make a decision whether the die or coins are biased?
NOTE: This stage students have covered Normal, Binomial, Uniform and Skewed distributions
Natural Variability – Exercise D
How do you make a decision whether the die or coins are biased?
Students are often divided on which ones are dodgy and which are fair. The simple
conclusion is that it can be difficult to determine. However, using statistical techniques we
can measure accurately which results are very unlikely once we take into account Natural
Variability (i.e. Randomness)
Implications for Teaching
(From Thinking and Reasoning Data and Chance, NCTM, 2006)
• Include ‘variability’ as a central issue.
– Explore ‘Natural Variability’
• Explore understanding of ‘center’
– What do we mean by ‘average’?
• Make connections between sample and populations.
– Understand the variability of a sample and how this effects Sampling Distributions

similar documents