This method is really more of a diagnostic or intuitive approach. The
results will certainly be possible (within the bounds), but the
assumptions are more intutive -- that the place on the tomography line
where a point probably is is a function of how close all the other lines
are to each point on a line. So everything depends entirely on whether
that assumption is reasonable in your setting. Note that the assumption
is 'local' in the sense that it applies to each line separately, rather
than 'global', such as in the regular parametric version of EI that fits a
specific model to the entire set of data. that is what makes it
nonparametric.
Gary
: Gary King, King(a)Harvard.Edu
http://GKing.Harvard.Edu :
: Center for Basic Research Direct (617) 495-2027 :
: in the Social Sciences Assistant (617) 495-9271 :
: 34 Kirkland Street, Rm. 2 HU-MIT DC (617) 495-4734 :
: Harvard U, Cambridge, MA 02138 eFax (928) 832-7022 :
On Wed, 10 Jul 2002, Linda Moore wrote:
Thanks for the responses to my earlier posted message
re setting of
priors.
I am still working on my gender data (looking at gender, turnout and
voting outcomes in NZ elections 1893-1954)- though take on board that
the limited variation between the groups may reduce the reliability of
existing ecological methods such as EI. As Stephen Voss suggested, I
certainly did find that the Palmquist inflation factor was very large,
and the standard errors were relatively high (eg:0.0147 for estimated
Female turnout of 0.9014).
What I have found since then is that the non-parametric version of EzI
seems to be giving good results where I can compare them with the truth.
eg: for the 1919 election the true values are
Female aggregate turnout=0.792 (min 0.655, max=0.882)
Male aggregate turnout=0.750 (min 0.539, max 0.871)
and the estimates using the non-parametric version with defaults are
BetaB (female) = 0.7760, 0.0166
BetaW (male) = 0.7661, 0.0157
I can slightly improve these estimates by shifting the defaults.
My question is - has anyone else used the non-parametric version, and/or
have any comments on its advantages and shortcomings?
Once again, any pointers would be much appreciated as I am not a
statistician and new to the area.
Thank you.
Linda
Linda Moore
Masters Student
History Department
University of Canterbury
Christchurch
NEW ZEALAND
lmm72(a)student.canterbury.ac.nz
-
ei mailing list served by Harvard-MIT Data Center
List Address: ei(a)latte.harvard.edu
Subscribe/Unsubscribe:
http://lists.hmdc.harvard.edu/listis.cgi?info=ei
-
ei mailing list served by Harvard-MIT Data Center
List Address: ei(a)latte.harvard.edu
Subscribe/Unsubscribe:
http://lists.hmdc.harvard.edu/listis.cgi?info=ei