gravitas numbers

[Grant Barrett]

David, as Ben points out, those numbers were from a Factiva search. I
did indeed explain to the reporter that Factiva's newspaper index may
have grown if it added new newspapers over the years we were
measuring--I did not expect him to use the numbers, only to report
that they tended to show an upward trend and two spikes. The trend
and the 2000 election spike is more apparent, by the way, if you
include the 1999 data, which only scores 1573 hits.

The proper way to do such data-gathering would have been to search a
set group of newspapers over that same period of time. So if we
search Factiva for just the New York Times, LA Times, Chicago
Tribune, Miami Herald, and Boston Globe--all which are indexed for
1999-2005--we get the same general upward trend and spikes in
election years as reported in the article:

1999 84
2000 169
2001 132
2002 143
2003 192
2004 236
2005 to date 197 (which will be 231 by Dec. 31 if it continues to be
used at the same rate)

Also interesting is to do the same numbers on television news
transcripts at LexisNexis. It also shows the same trends and peaks,
although more pronounced. All the transcript sources have data in
LexisNexis for the entire 1999-2005 period. This is likely to include
any transcription errors.

1999 76
2000 314
2001 45
2002 58
2003 66
2004 194
2005 74 (trending to 87 by Dec. 31)

I'm not sure how many false positives you're going to get on a word
like "gravitas." In any case, for a non-critical count like this one,
couldn't it be accepted that the error rate would be relatively the
same across the entire 1999-2005 period?

Grant Barrett
gbarrett at worldnewyork.org
http://www.doubletongued.org/