----- Original Message -----
From: Alan Statistics
To: AllStat List
Sent: Monday, March 15, 2010 4:09 PM
Subject: Spiked (too angular to be normal) distribution of differences between two sets of measurements of same objects
With Respect
For those who haven't realized, this is another silly amateur question.
Looking at the distribution of a sample of cases by differences between two measures of their subsequent citation, I found it was almost flat except for the central seventh which rose in a spike.
If there were no correlation between the two measures, I would expect a normal or pyramidal distribution of differences, at least from extrapolating from the perfect horizontal line of total negative correlation and perfect vertical line of total positive correlation. But this distribution seems a fusion of rather than a mean between the two extremes.
Now I have found the same type of distribution in the differences for both homicide and infant mortality rates given for a sample of countries between the 2009 Britannica Yearbook and the 1994 Yearbook.
Can anyone provide me with an explanation, or a reference to an explanation, of the cause, properties, and implications of this distribution of differences?
Yours Sincerely,
Alan E. Dunne
You may leave the list at any time by sending the command
SIGNOFF allstat
to [log in to unmask], leaving the subject line blank.
|