From: stephanh@ix.netcom.com (Stephan Heumann)
Newsgroups: alt.hemp,alt.hemp.politics,talk.politics.drugs,alt.politics.democrats.d
Subject: Re: Clue to Clinton: 85% ... Statistical remarks.
Date: 25 May 1995 08:24:40 GMT
Message-ID: <3q1es8$sh6@ixnews3.ix.netcom.com>

gal2@kimbark.uchicago.edu (Jacob Galley) writes: 
>
>
>Even if the President read his mail, your survey shows nothing without
>proper statistical analysis (and even then, very little).
>
>I have data here from the National Opinion Research Center's General
>Social Survey on the legalization of marijuana by age group and
>education level.  The GSS is a multistage, stratified probability
>sample of English speaking adults living in US households.  That is,
>the data is collected by interviewing randomly selected adults in
>randomly selected houses on randomly selected streets in randomly
>selected towns.  The survey has been administered since 1972, and is
>supported by NSF.  The table below uses data from the 1990-94 data
>sets.
>
>
>Q: Do you think that the use of marijuana should be made legal or not?
>
>		      Percent who approve of legalization
>
>AGE		NO BACHELOR'S  (100%=)		BACHELOR'S  (100%=)
>
>18-34		    25.2%	(1157)		   22.8%      (307)
>35-64		    21.5%	(1757)		   26.1%      (631)
>65+		    11.2%	 (752)		   21.4%      (126)
>
>All		    20.5%	(3666)		   24.6%     (1064)
>
>
>The numbers in parentheses show how many people were sampled in each
>category.  For the non-college-graduates, p < .000005, meaning that
>the trend across age groups is highly significant.  For the college
>graduates, p = .3, so no significant relationship between age and
>legalization preference exists.
>

Jake-
Thanks for the valuable (if discouraging) statistics. I can't dispute
these results when the survey question is posed in terms of point-blank
legalization, with no specific, coherent policy proposed. That's why
the familiar model of the wine laws is useful.

And granted, a self-selected sample can be highly biased (or weighted
by motivation on either side), so the fact that Parade magazine's
phone-in poll last year (with a $1 900 number) found 76% agreement that
"marijuana should be as legal as wine" adds no credence to my result.

But at about the same time, I conducted a *phone-out* poll using random
Palo Alto phone numbers and also came up with 76% agreement (in a
sample of 50) that "the marijuana laws should be reformed to be no more
restrictive than the wine laws."  OK, it's a well-educated population.

Still, the voters are also a self-selected, motivation-weighted sample!

One academic note on "significance" of age correlation: evidently the
p-values cited are computed for a linear model. With such large and
unequal age cells, this seems like a highly questionable procedure. But
the non-linearity of levels of pro-legalization sentiment with age
among college graduates doesn't bother me, so I'll pass on questioning
it. Except to lament the fact that recent graduates are (purportedly)
even more anti-legalization than non-graduates in their age bracket.

It _really_ makes me worry about the state of higher education today.
It seems that lower education (e.g., DARE) has gained the upper hand.

(/..\)
-Stephan