24 May
2010
24 May
'10
7:31 p.m.
When using the fn=NULL option in setx, how does one calculate the
standard error of the resulting expected values? Is it just "sd" from
the output divided by the square root of the number of cases in the
data? Or something else? I'm confused because "sd" for conditional
prediction is much larger than "sd" for a unconditional prediction (like
setting all vars to their sample means), even though the former seems
grounded in more real information.
Thanks,
Eric
Eric McGhee
Research Fellow
PUBLIC POLICY
INSTITUTE OF CALIFORNIA
500 Washington Street, Suite 600
San Francisco, CA 94111
tel 415 291 4439
fax 415 291 4401
web www.ppic.org <http://www.ppic.org>
Any opinions expressed in this message are those of the author alone and
do not necessarily reflect any position of the Public Policy Institute
of California.
25 May
25 May
9:11 a.m.
Standard errors are all calculated based on the simulations.
Unconditional prediction averages over all observations in the sample and
so if the sample is heterogenous (in terms of explanatory variables), the
variance of the resulting predictions can be high.
Kosuke
--
Department of Politics
Princeton University
http://imai.princeton.edu
On Mon, 24 May 2010, Eric McGhee wrote:
When using the fn=NULL option in setx, how does one
calculate the
standard error of the resulting expected values? Is it just "sd" from
the output divided by the square root of the number of cases in the
data? Or something else? I'm confused because "sd" for conditional
prediction is much larger than "sd" for a unconditional prediction (like
setting all vars to their sample means), even though the former seems
grounded in more real information.
Thanks,
Eric
Eric McGhee
Research Fellow
PUBLIC POLICY
INSTITUTE OF CALIFORNIA
500 Washington Street, Suite 600
San Francisco, CA 94111
tel 415 291 4439
fax 415 291 4401
web www.ppic.org <http://www.ppic.org>
Any opinions expressed in this message are those of the author alone and
do not necessarily reflect any position of the Public Policy Institute
of California.
-
Zelig Mailing List, served by Harvard-MIT Data Center
Send messages: zelig(a)lists.gking.harvard.edu
[un]subscribe Options: http://lists.gking.harvard.edu/?info=zelig
Zelig program information: http://gking.harvard.edu/zelig/
12:39 p.m.
Let me follow up briefly. When I run these commands
z.out <- zelig(superbd2 ~ age+income+educ+pid7+ideo5+
ageXginfo+incXginfo+educXginfo+pidXginfo+ideoXginfo+ginfo,
model="logit", data=ppicjan10)
x.out <- setx(z.out, fn=NULL)
s.out <- sim(z.out, x=x.out)
where superbd2 is dichotomous with a mean of 0.504, the mean expected
value is 0.504 (makes sense) and the "sd" is 0.1446. I'm assuming
"sd"
means standard deviation, not standard error? So I need to divide this
number by sqrt(n) to get the standard error for significance
calculations? Pardon if this question is simple-minded.
Best,
Eric
Eric McGhee | Research Fellow | PPIC | 415-291-4439
Any opinions expressed in this message are those of the author alone and
do not necessarily reflect any position of the Public Policy Institute
of California.
-----Original Message-----
From: Kosuke Imai [mailto:kimai@Princeton.Edu]
Sent: Tuesday, May 25, 2010 6:11 AM
To: Eric McGhee
Cc: zelig(a)lists.gking.harvard.edu
Subject: Re: [zelig] simple question
Standard errors are all calculated based on the simulations.
Unconditional prediction averages over all observations in the sample
and
so if the sample is heterogenous (in terms of explanatory variables),
the
variance of the resulting predictions can be high.
Kosuke
--
Department of Politics
Princeton University
http://imai.princeton.edu
On Mon, 24 May 2010, Eric McGhee wrote:
When using the fn=NULL option in setx, how does one
calculate the
standard error of the resulting expected values? Is it just "sd" from
the output divided by the square root of the number of cases in the
data? Or something else? I'm confused because "sd" for conditional
prediction is much larger than "sd" for a unconditional prediction
(like
setting all vars to their sample means), even though
the former seems
grounded in more real information.
Thanks,
Eric
Eric McGhee
Research Fellow
PUBLIC POLICY
INSTITUTE OF CALIFORNIA
500 Washington Street, Suite 600
San Francisco, CA 94111
tel 415 291 4439
fax 415 291 4401
web www.ppic.org <http://www.ppic.org>
Any opinions expressed in this message are those of the author alone
and
do not necessarily reflect any position of the Public
Policy Institute
of California.
-
Zelig Mailing List, served by Harvard-MIT Data Center
Send messages: zelig(a)lists.gking.harvard.edu
[un]subscribe Options: http://lists.gking.harvard.edu/?info=zelig
Zelig program information: http://gking.harvard.edu/zelig/
1:51 p.m.
"sd" is the standard deviation of posterior distribution, which is
equivalent to standard error asymptotically. So, you can interpret it as
standard error.
Kosuke
--
Department of Politics
Princeton University
http://imai.princeton.edu
On Tue, 25 May 2010, Eric McGhee wrote:
Let me follow up briefly. When I run these commands
z.out <- zelig(superbd2 ~ age+income+educ+pid7+ideo5+
ageXginfo+incXginfo+educXginfo+pidXginfo+ideoXginfo+ginfo,
model="logit", data=ppicjan10)
x.out <- setx(z.out, fn=NULL)
s.out <- sim(z.out, x=x.out)
where superbd2 is dichotomous with a mean of 0.504, the mean expected
value is 0.504 (makes sense) and the "sd" is 0.1446. I'm assuming
"sd"
means standard deviation, not standard error? So I need to divide this
number by sqrt(n) to get the standard error for significance
calculations? Pardon if this question is simple-minded.
Best,
Eric
Eric McGhee | Research Fellow | PPIC | 415-291-4439
Any opinions expressed in this message are those of the author alone and
do not necessarily reflect any position of the Public Policy Institute
of California.
-----Original Message-----
From: Kosuke Imai [mailto:kimai@Princeton.Edu]
Sent: Tuesday, May 25, 2010 6:11 AM
To: Eric McGhee
Cc: zelig(a)lists.gking.harvard.edu
Subject: Re: [zelig] simple question
Standard errors are all calculated based on the simulations.
Unconditional prediction averages over all observations in the sample
and
so if the sample is heterogenous (in terms of explanatory variables),
the
variance of the resulting predictions can be high.
Kosuke
-
Zelig Mailing List, served by Harvard-MIT Data Center
Send messages: zelig(a)lists.gking.harvard.edu
[un]subscribe Options: http://lists.gking.harvard.edu/?info=zelig
Zelig program information: http://gking.harvard.edu/zelig/
5108
days inactive
5109
days old
3 comments
2 participants
participants (2)
-
Eric McGhee -
Kosuke Imai