The Lincoln Plawg - the blog with footnotes

Politics and law from a British perspective (hence Politics LAW BloG): ''People who like this sort of thing...'' as the Great Man said

This page is powered by Blogger. Isn't yours?
Wednesday, January 19, 2005
 

New York Times: niggardly with niggardly


I'm listening to a less-annoying-than-usual Al Franken interview - with Peter Beinart - which alludes to a Times piece calling USG stingy with development aid to Third World countries.

The brain whirs: for stingy, I'm thinking niggardly - on which Washington mayor Anthony Williams had such fun a few years ago.

So, I go to the Times search page: since 1996, the paper has used the word a mere 35 times.

And the last time it was used was in March 2002 - the piece is nowhere online, that I can see.

The usage numbers:

2002 - 1
2001 - 2
2000 - 5
1999 - 16
1998 - 5
1997 - 3
1996 - 2

The numbers are skewed by the 1999 controversy over Williams' aide David Howard's use of the word.

Is the fact that 2003 is the first of recent years in which niggardly has not appeared in the Times an indication of creeping censorship? Or had censorship set in well and truly by 1996 to produce so - niggardly - a number as two uses in one year?

What about the Post? The archives are divided into 1987 to date and 1877-86: the stats for recent years are these:

2003 - 1
2002 - 1
2001 - 1
2000 - 6
1999 - 38
1998 - 3
1997 - 2
1996 - 3
1995 - 2
1994 - 4
1993 - 3
1992 - 4
1991 - 3
1990 - 3
1989 - 13
1988 - 3
1987 - 9
1986 - 9
1985 - 4
1984 - 8
1983 - 8
1982 - 10
1981 - 10

My guess is that there is nothing statistically significant there.

What about a decade before the PC nonsense started: say, the 1930s:

1930 - 9
1931 - 13
1932 - 12
1933 - 7
1934 - 12
1935 - 18
1936 - 11
1937 - 6
1938 - 15
1939 - 17
1940 - 9

A statistically significant difference between the 1930s and the last decade? Factor in the much greater size of today's papers: a distinct possibility.


MORE

With some trepidation, I offer some statistical analysis: using the t test - an online calculator does the work! - to compare the number for use of niggardly by the Post in the 1930s with the 11 most recent years (1993-2003), one finds that the difference between the samples is significant only at the 90% level - ie, not significant at all.

Using the Wilcoxon Two Sample test, however, the difference is staggeringly significant - at the 99.9% level (where 95% is generally thought sufficient).

My guess is that the large count for 1999 skews the distribution of the data in the later sample - and the t test only works with normally distributed (bell curve) data. Whereas the Wilcoxon test is, as they say, nonparametric - works regardless of distribution.

Plus, in this case, visual inspection and common sense concludes a huge difference between the samples.


|
free website counter Weblog Commenting and Trackback by HaloScan.com