Like the binomial logit, you should do the simulation. The coefficients
are distributed as a multivariate normal distribution. Calculate fitted
values and first differences using the simulation.
Kosuke
On Tue, 29 Apr 2003, Stanislav Markus wrote:
Trying to decipher the multinomial logit coefficients,
I wonder if
somebody can help with the interpretation...
Thank you,
Stan
-----Original Message-----
From: gov2001-l-admin(a)fas.harvard.edu
[mailto:gov2001-l-admin@fas.harvard.edu] On Behalf Of Kosuke Imai
Sent: Tuesday, April 29, 2003 9:15 PM
To: Anna Lorien Nelson
Cc: gov2001-l(a)fas.harvard.edu
Subject: Re: [gov2001-l] standard errors of interaction terms
Use the simulations just like we did in the class.
Kosuke
On Tue, 29 Apr 2003, Anna Lorien Nelson wrote:
When you run a linear regression that includes
interaction terms, how
do
you interpret the standard errors?
For instance, we have two dummy variables, one that indicates the
presence
of a popular initiative on a ballot and another
that indicates whether
a
presidential race was held that year. So we get
estimated
coefficients
and standard errors for:
initiative
presrace
initiative*presrace
When initiative and presrace both = 1 (both present on the ballot),
the
total predicted effect of the presence of an
initiative equals the
coefficient on initiative plus the coefficient on initiative*presrace.
What is the standard error for this predicted effect? Do we add
together
the standard errors for initiative +
initiative*presrace, or do
something
else?
Thanks!
Anna
_______________________________________________
gov2001-l mailing list
gov2001-l(a)fas.harvard.edu
http://www.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)fas.harvard.edu
http://www.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)fas.harvard.edu
http://www.fas.harvard.edu/mailman/listinfo/gov2001-l