6 May
2003
6 May
'03
9:40 a.m.
Dear all,
Quick question: does calculating a quantity of interest on 5 datasets and then
averaging the answer produce the same answer as combining the 5 into 1 dataset
and then doing the calculations?
Many thanks.
Best,
Dan
--
PhD Candidate in Government
Proctor
Harvard University
(617) 493-8402
6 May
6 May
9:43 a.m.
nope. you have to run it separately on the 5 datasets.
Gary
On Tue, 6 May 2003 dhopkins(a)fas.harvard.edu wrote:
Dear all,
Quick question: does calculating a quantity of interest on 5 datasets and then
averaging the answer produce the same answer as combining the 5 into 1 dataset
and then doing the calculations?
Many thanks.
Best,
Dan
9:46 a.m.
but you know how to program of course, so write a function that does
whatever you want 5 times. once you've written the function, its as easy
to use the imputed data sets as using one!
Gary
On Tue, 6 May 2003, Gary King wrote:
nope. you have to run it separately on the 5 datasets.
Gary
On Tue, 6 May 2003 dhopkins(a)fas.harvard.edu wrote:
Dear all,
Quick question: does calculating a quantity of interest on 5 datasets and then
averaging the answer produce the same answer as combining the 5 into 1 dataset
and then doing the calculations?
Many thanks.
Best,
Dan
_______________________________________________
gov2001-l mailing list
gov2001-l(a)fas.harvard.edu
http://www.fas.harvard.edu/mailman/listinfo/gov2001-l
8:35 p.m.
New subject: [gov2001-l] AIC in model selection
How valuable or interesting is AIC - as compared, say, to individual
coefficients' size or t-statistics - in selecting among the models?
Thank you,
Stan
****************************
Stanislav Markus
Ph.D. Candidate
Harvard University
Department of Government
e: smarkus(a)fas.harvard.edu
t: 617.513.5407
10:18 p.m.
New subject: [gov2001-l] AIC in model selection
i'd compute quantities of interest.
then make sure the model fits the data.
once you've done those things, you don't need AIC. AIC is a way of
choosing among a set of models. but its no magic bullet. its basically
R^2 minus a penalty for every extra variable you include. just like
adjusted R^2.
Gary
On Tue, 6 May 2003, Stanislav Markus wrote:
How valuable or interesting is AIC - as compared, say,
to individual
coefficients' size or t-statistics - in selecting among the models?
Thank you,
Stan
****************************
Stanislav Markus
Ph.D. Candidate
Harvard University
Department of Government
e: smarkus(a)fas.harvard.edu
t: 617.513.5407
_______________________________________________
gov2001-l mailing list
gov2001-l(a)fas.harvard.edu
http://www.fas.harvard.edu/mailman/listinfo/gov2001-l
7683
days inactive
7684
days old
gov2001@lists.gking.harvard.edu
4 comments
3 participants
participants (3)
-
dhopkins＠fas.harvard.edu -
Gary King -
smarkus＠fas.harvard.edu