On Mon, 08 Apr 2013 08:55:12 -0700, Celia Reaves wrote:
For those who teach statistics and the various ways of displaying data,
a current ad by Prudential may be interesting. It shows a real-life example
of a bar graph.
https://www.youtube.com/watch?v=C3qj88J7-jA
Oh man, whenever this ad comes on TV I feel like throwing my shoe
at the TV. Problems with Daniel Gilbert's ("Mr. Happiness") include
giving point estimates for life expectancy instead of interval estimates
and confusing matters by talking about life expectancy when his DAD
was a kid instead of what it is today.
He then changes the nature of the conversation when he says
"to show how things have changed", we asked people "how old is
the oldest person you know". Now this has no bearing on what
Gilbert said first because he did not provide comparable info for
when his father was a kid. Instead, he should have provided what
the average life expectancy is now which, admittedly, is misleading
since different groups have different life expectancies. Similarly,
asking "how old is the oldest person you know" is misleading because:
(1) Different groups (based on gender, race/ethnicity, SES, occupation,
etc.) will have different oldest ages. If Gilbert was using a stratified
random sampling frame incorporating the aforementioned factor, then
the oldest ages presented might be meaningful. Otherwise, no, the
values shown are not representative of the population(s) being
addressed.
(2) One obvious problem is whether the same oldest person is being
counted twice or more often. Technically, if several people nominate
the same old person, that should only count once. Clearly, members
of a family probably all know the same old person and their vote
should only count once. Alas, Gilbert provides no method section
and one cannot determine how this problem was dealt with. ;-)
(3) Technically, what Gilbert is presenting is a dot histogram and not
a bar chart or ordinary histogram.
(4) Gilbert makes the claim that "people are now living longer" which
is probably true (though he never says "living longer" relative to what
group or time frame or context) but this statement is not necessarily
supported by the dot histogram he presents. He needs to have a
comparable dot histogram for different time periods and/or groups
and/or etc. Perhaps this is just pilot data and maybe he'll get funding
for a more serious research project (perhaps from some insurance
company ;-).
-Mike Palij
New York University
[email protected]
---
You are currently subscribed to tips as: [email protected].
To unsubscribe click here:
http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5&n=T&l=tips&o=24857
or send a blank email to
leave-24857-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
