Transcript Statistics in Business

Statistics in Business
Phillip E Pfeifer
Statistics in Business
Phillip E Pfeifer
Darden Graduate Business School
1980 to present
Georgia Tech
1979
Ted Pfeifer
Stevens Tech 2010
Max Pfeifer
ND 2012
Thomas Jefferson Had Girls
Randolph
Jefferson
17551815
Sally
Hemings
1773-1835
Thomas Jefferson
1743-1826
Eston Hemings
1806-1856
Pointing to Randolph Jefferson, we have
evidence he was at Monticello nine
months before Sally gave birth to
Eston—the DNA match. We got
evidence that he used to go party and
socialize with the slaves late at night…
What was the most important evidence
pointing to Randolph and not Thomas?
He said, “Look, the DNA match was to a
male child, Eston Hemings. Randolph
Jefferson had six male children with two
wives. Thomas Jefferson had all female
children, except for an infant who died.
Who is more likely to father a male
child, Randolph or Thomas?”
I thought that was a brilliant point.
David A. Maurer, “In Defense of a
Reputation,” Daily Progress, January 18,
2010.
I have some data….
• This data set is
great.
• These data are
great.
• The first four
variables are
categorical.
• The last variable
(#children) is
numeric.
• #children is
discrete (not
continuous).
• There are 700,300
elements (families)
in the data set.
• n=700,300
Expected vs Actual 4-child Family
counts by NumberofFemales
Famsize
4
Column Labels
Count of ID
Expected counts
Probability
0
2699
1975
0.0625
1
2
8392 11030
7899 11848
0.25 0.375
3
4 Grand Total
7435 2039
31595
7899 1975
31595
0.25 0.063
1
The actual Counts for 0 and 4 are too high (and significantly so…says pfeifer)
The actual Count for 2 is too low
Evidence that either the coin flip model is not correct .. Or….