Chance News 95: Difference between revisions

From ChanceWiki
Jump to navigation Jump to search
Line 110: Line 110:
This gives an example of Simpson's Paradox.  In both the north and the south, the bill had stronger support among Democrats than Republicans.  However, aggregating over region leads to the first table, which gives the reverses the direction of the association.  The explanation is that a larger proportion of Democrats came from the south (they outnumbered Republicans 91 to 11 in the House and 21 to 1 in the Senate), where support for the Civil Rights Act was weaker.
This gives an example of Simpson's Paradox.  In both the north and the south, the bill had stronger support among Democrats than Republicans.  However, aggregating over region leads to the first table, which gives the reverses the direction of the association.  The explanation is that a larger proportion of Democrats came from the south (they outnumbered Republicans 91 to 11 in the House and 21 to 1 in the Senate), where support for the Civil Rights Act was weaker.


['''Note'''.  In the original ''Guardian'' article, the third table gives the Dem/Union count in the House as 144, which leaves the Aye total one vote short of the earlier tables.  Nick Horton alertly noticed this and wrote to the author, who subsequently identified the missing vote as Pennsylvania Democrat whose party affiliation was listed as "unknown" in one of the voting databases.  Thanks to Nick for communicating the correction to the Isolated Statisticians list.]
['''Note'''.  In the original ''Guardian'' article, the third table gives the Dem/Union count in the House as 144, which leaves the Aye total one vote short of the earlier tables.  Nick Horton alertly noticed this and wrote to the author, who subsequently identified the missing vote as a Pennsylvania Democrat whose party affiliation was listed as "unknown" in one of the voting databases.  Thanks to Nick for communicating the correction to the Isolated Statisticians list.]

Revision as of 20:08, 6 September 2013

Quotations

“It is worth dwelling for a moment on Egon [son of Karl] Pearson’s first-year lecture course …. [H]e was an inspirational teacher …. What was the reason for his success? [H]e was not a teacher who ladled out cookery-book recipes; rather he always seemed in his lectures to be working through and exploring problems with the class. He would wander down enticing dead-ends, but return to seek alternatives again and again until a satisfactory approach had been established. The result was that students acquired a questioning approach, not a compartmentalized approach whereby one problem was allocated to a 2 x 2 table, the next to multiple linear regression, etc.”

“What is Statistics?” by David J. Bartholomew

(1994 presidential address to the RSS)

Journal of the Royal Statistical Society, 1995

Submitted by Margaret Cibes


“In my own field of flood risks, a talented statistician declared: ‘It is also true that for extremely rare events, correct uncertainty estimates may lead us to conclude that we know virtually nothing. This is not such a bad thing. If we really know nothing we should say so!’”

Letter to editor, Significance magazine, June 2013

Submitted by Margaret Cibes

Forsooth

TopTenWorstGraphs.jpg
“Figure 2. Q-Q plots of Z scores for telomeric interval-length differences.”


Cited as #8 of “The top ten worst graphs”
from “Ethnicity and Human Genetic Linkage Maps”, American Journal of Human Genetics, February 2005


EconGraph.jpg
“What Is Economics Good For?”
The New York Times, August 24, 2013
ConfVar.png
Significance magazine, March 2011

Submitted by Margaret Cibes and James Greenwood


Weeding wedding invitation lists

“GUESTimation: Breaking the deadlock on wedding guest lists”
by Damjan Vukcevic, Significance, August 2013

Winner of the second annual Young Writers Competition, this article describes the process the author used to narrow his initial wedding guest list down to a number that his venue could accommodate. The process included grouping potential invitees (e.g., families), ranking them for their probabilities of attending if invited (to Australia), and using a probability distribution of the number of attendees to get a confidence interval of attendees. He also discusses his independence assumption and the consequence of using or not using it.

Submitted by Margaret Cibes

Civil rights and Simpson's paradox

Mary Parker sent this link to the Isolated Statisticians list:

Were Republicans really the party of civil rights in the 1960s?
by Harry J. Enten, Guardian, 28 August 2013

The 50th anniversary of the March on Washington and Martin Luther King's famous "I have a dream speech" have been commemorated in much recent news coverage. In light of some Republican claims that their record on civil rights compares favorably to Democrats, the Guardian takes a statistical look at the voting record on the landmark Civil RIghts Act of 1964.

They present 3 tables, which are reproduced below. The first breaks out the vote by party in the House of Representatives and the Senate, indicating that the bill had greater support among Republicans.

1964 Civil Rights Act Senate Version Ayes
Democrats Republicans
House 153 of 244 (63%) 136 of 171 (80%)
Senate 46 of 67 (69%) 27 of 33 (82%)

The second table takes into account the history of the Civil War, separating out the 11 southern states that formed the Confederacy (note the remaining 39 states are classified as "Union", but of course not all of these current states existed in the 1860s).

1964 Civil Rights Act Senate Version Ayes
Union Confederacy
House 281 of 313 (90%) 8 of 102 (8%)
Senate 72 of 78 (92%) 1 of 22 (5%)

Observing that political party and geography both matter, the Guardian's third table accounts for both:

1964 Civil Rights Act Senate Version Ayes
Dem/Union Rep/Union Dem/Confed Rep/Confed
House 145 of 152 (95%) 137 of 161 (85%) 8 of 91 (9%) 0 of 11 (0%)
Senate 45 of 46 (98%) 27 of 32 (84%) 1 of 21 (5%) 0 of 1 (0%)

This gives an example of Simpson's Paradox. In both the north and the south, the bill had stronger support among Democrats than Republicans. However, aggregating over region leads to the first table, which gives the reverses the direction of the association. The explanation is that a larger proportion of Democrats came from the south (they outnumbered Republicans 91 to 11 in the House and 21 to 1 in the Senate), where support for the Civil Rights Act was weaker.

[Note. In the original Guardian article, the third table gives the Dem/Union count in the House as 144, which leaves the Aye total one vote short of the earlier tables. Nick Horton alertly noticed this and wrote to the author, who subsequently identified the missing vote as a Pennsylvania Democrat whose party affiliation was listed as "unknown" in one of the voting databases. Thanks to Nick for communicating the correction to the Isolated Statisticians list.]