Date: Wed, 2 Mar 2005 09:46:43 +0100
Reply-To: Spousta Jan <JSpousta@CSAS.CZ>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Spousta Jan <JSpousta@CSAS.CZ>
Subject: Re: ics
Content-Type: text/plain; charset="koi8-r"
Hi Anton,
Just trying to think a bit further:
"Is it possible to affirm that there are greater differences at certain questions between generations than within ones?"
"Differences" in beliefs - is it perhaps a measure of unanimity (or of its opposite, diversity of views)? If yes, then the simple row percent will work as the measure of homogemeity in a category - see how near is it to 1/2 = complete diversity.
And you can construct the confidence intervals of the percentages in a standard manner.
Or, if you will be more scientific, study entropy of the responses or conditional probabilities p(they agree | they are from the same generation) vs. p(they agree | they are from different generations). But I doubt whether this is worth studying in such a simple case.
HTH
Jan
-----Original Message-----
From: SPSSX(r) Discussion [mailto:SPSSX-L@LISTSERV.UGA.EDU]On Behalf Of Anton Balabanov
Sent: Tuesday, March 01, 2005 3:02 PM
To: SPSSX-L@LISTSERV.UGA.EDU
Subject: ics
Dear Listers,
my question is rather statistical one (if it does make sense itself :).
My father doing a sociological survey trying to assess knowledge and
opinions of three generations of Russian people regarding World War II and,
in particular, Great Patriotic War of Soviet people.
For some questions from the questionnaire it make sense to construct
contingency tables like this:
Sons Yes 50
Fathers Yes 60
Grandpas Yes 95
Sons No 33
Fathers No 55
Grandpas No 40
Father has asked me to help with a) whether the generations and answers are
statistically independent and b) is it possible to affirm that there are
greater differences at curtain questions between generations than within
ones.
Regarding a) SPSS provides enough measures in CROSSTABS to answer;
But I am in complete stupor with b). Is there any metric to estimate "total
heterogeneity" of contingency table and than "heterogeneity due to rows" and
"due to columns"?
Kind regards,
Anton Balabanov
Assistant of the Dept. of Applied Statistics
University of Nizhny Novgorod, Russia
