9 Dec
2004
9 Dec
'04
8:18 a.m.
Peeps: Just a reminder -- Both sections will be held in Science Center
316 today. See you there! Olivia.
15 Dec
15 Dec
9:12 a.m.
New subject: [gov2001-l] mix package
is 'imp.mix' the right command to use in the mix package for the
imputations? it's the only imputation type command i cound find there.
'imp.mix' apparently requires pre-processing with 'da.mix' or
'dabipf.mix'
do we need to use any of those too?
thanks!
in da mix,
j.
9:23 a.m.
New subject: [gov2001-l] mix package
You should follow the steps in the example at the bottom of
help(imp.mix)...so yes, you do need to do some pre-processing before you
use imp.mix().
Also, note that I forgot to mention something in section last week: If we
fix tau1 = 0, we need to fix tau2 > 0. So you need to exp(par[k+1]) it to
contstrain it to be greater than zero, but I'm sure that everyone has
figured that out by now... 8)
Olivia.
On Wed, 15 Dec 2004, Jens Hainmueller wrote:
is 'imp.mix' the right command to use in the
mix package for the
imputations? it's the only imputation type command i cound find there.
'imp.mix' apparently requires pre-processing with 'da.mix' or
'dabipf.mix'
do we need to use any of those too?
thanks!
in da mix,
j.
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
1:20 p.m.
New subject: AW: [gov2001-l] mix package
http://cran.r-project.org/doc/packages/mix.pdf
contains a detailed and very helpful documentation of the mix package.
best,
j.
-----Ursprungliche Nachricht-----
Von: gov2001-l-bounces(a)lists.fas.harvard.edu
[mailto:gov2001-l-bounces@lists.fas.harvard.edu]
Gesendet: Wednesday, December 15, 2004 2:23 PM
An: gov2001-l(a)lists.fas.harvard.edu
Betreff: Re: [gov2001-l] mix package
You should follow the steps in the example at the bottom of
help(imp.mix)...so yes, you do need to do some pre-processing before you
use imp.mix().
Also, note that I forgot to mention something in section last week: If we
fix tau1 = 0, we need to fix tau2 > 0. So you need to exp(par[k+1]) it to
contstrain it to be greater than zero, but I'm sure that everyone has
figured that out by now... 8)
Olivia.
On Wed, 15 Dec 2004, Jens Hainmueller wrote:
is 'imp.mix' the right command to use
in the mix package for the
imputations? it's the only imputation type command i cound find there.
'imp.mix' apparently requires pre-processing with 'da.mix' or
'dabipf.mix'
do we need to use any of those too?
thanks!
in da mix,
j.
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
16 Dec
16 Dec
5:24 a.m.
New subject: [gov2001-l] Re: ordered probit coefficients
Hi -
I know that we are not supposed to be interested in the raw coefficient
estimates from something like ordered probit, but how should we think
about the fact that the coefficient estimates from Zelig are of the
opposite sign while the intercept/threshold estimates are of the same sign
as what we are getting from our own code (and from what you get using
polr() in the MASS package)? What if the substantive question of interest
is the underlying unobserved variable?
Cheers
Mike
5:49 a.m.
New subject: [gov2001-l] Re: ordered probit coefficients
Excellent questions, mike.
1) Just so everyone is clear on this: polr() is for ordinal
*logit* not ordinal probit.
2) The difference is one in parameterization. Remember how we
fixed tau1 = 0? If you fix tau1 to a different value, you will
get a different answer for the second intercept term (tau2) and
different signs on the coefficients (beta). As long as you use
the same parameterization to calculate the quantities of
interest, the predicted probability of falling into each bin
will be invariant to reparameterization of the systematic
component.
3) The substantive question of interest can't be the unobserved
underlying variable (because the stochastic component is
multinomial). Thus, the expected value for this model is the
predicted probability that an observation i falls into each bin.
So let's say that you're interested in presidential approval.
Someone conducts a survey asking: "The president is doing an
effective job. Strongly agree (1), agree (2), neutral (3),
disagree (4), strongly disagree (5)." We observe the number of
people in bins 1:5 and estimate the model and find beta and
tau2-tau5. If we were just interested in the unobserved
underlying distribution Y*, the quantity we would calculate is
x'beta = mu. Now what does mu mean? mu relative to what? It
has to be mu relative to the cut points.
Olivia
----- Original Message -----
From: "Michael Richard Kellermann" <kellerm(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Thursday, December 16, 2004 10:24 AM
Subject: [gov2001-l] Re: ordered probit coefficients
Hi -
I know that we are not supposed to be interested in the raw
coefficient
estimates from something like ordered probit, but how should
we think
about the fact that the coefficient estimates from Zelig are
of the
opposite sign while the intercept/threshold estimates are of
the same sign
as what we are getting from our own code (and from what you
get using
polr() in the MASS package)? What if the substantive question
of interest
is the underlying unobserved variable?
Cheers
Mike
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
6:20 a.m.
New subject: [gov2001-l] Re: ordered probit coefficients
On Thu, 16 Dec 2004, Olivia Lau wrote:
Excellent questions, mike.
1) Just so everyone is clear on this: polr() is for ordinal
*logit* not ordinal probit.
polr has an ordered probit option. Quoting from the polr help file:
The ordered factor which is observed is which bin
Y_i falls into with breakpoints
zeta_0 = -Inf < zeta_1 < ... < zeta_K = Inf
This leads to the model
logit P(Y <= k | x) = zeta_k - eta
with _logit_ replaced by _probit_ for a normal latent variable,
and eta being the linear predictor, a linear function of the
explanatory variables.
At any rate, it gives the same estimates as Stata (and as Zelig, except
for the sign change).
2) The difference is one in parameterization. Remember how we
fixed tau1 = 0? If you fix tau1 to a different value, you will
get a different answer for the second intercept term (tau2) and
different signs on the coefficients (beta). As long as you use
the same parameterization to calculate the quantities of
interest, the predicted probability of falling into each bin
will be invariant to reparameterization of the systematic
component.
I buy that; what I don't understand is why you gould get different signs
on the coefficients (beta) when you are getting the same estimates (signs
and magnitudes) on the taus.
3) The substantive question of interest can't be
the unobserved
underlying variable (because the stochastic component is
multinomial). Thus, the expected value for this model is the
predicted probability that an observation i falls into each bin.
So let's say that you're interested in presidential approval.
Someone conducts a survey asking: "The president is doing an
effective job. Strongly agree (1), agree (2), neutral (3),
disagree (4), strongly disagree (5)." We observe the number of
people in bins 1:5 and estimate the model and find beta and
tau2-tau5. If we were just interested in the unobserved
underlying distribution Y*, the quantity we would calculate is
x'beta = mu. Now what does mu mean? mu relative to what? It
has to be mu relative to the cut points.
This is what I don't understand. Say you were interested in the effects
of certain covariates on the heights of children. You don't collect the
data, and for some reason the researchers who did collect the data only
tell you which of five categories they are in - really short, kind of
short, average, kind of tall, really tall - and you don't know what the
cutpoints were that they used to divide the children into categories.
Maybe I'm missing something, but this would seem to be the kind of
question one might be interested in answering using something like ordered
probit. In this case, what bin you are in is of no substantive interest,
it is just an artifact of the data collection process. It is the mu that
we are interested in, even if we don't have it in meaningful units. Now,
if I estimate this model with zelig and the coefficient on (say) income is
negative and in polr it is positive, what is one supposed to think?
Cheers
Mike
Olivia
----- Original Message -----
From: "Michael Richard Kellermann" <kellerm(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Thursday, December 16, 2004 10:24 AM
Subject: [gov2001-l] Re: ordered probit coefficients
Hi -
I know that we are not supposed to be interested in the raw
coefficient
estimates from something like ordered probit, but how should
we think
about the fact that the coefficient estimates from Zelig are
of the
opposite sign while the intercept/threshold estimates are of
the same sign
as what we are getting from our own code (and from what you
get using
polr() in the MASS package)? What if the substantive question
of interest
is the underlying unobserved variable?
Cheers
Mike
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
6:47 a.m.
New subject: [gov2001-l] Re: ordered probit coefficients
2) The
difference is one in parameterization. Remember how we
fixed tau1 = 0? If you fix tau1 to a different value, you will
get a different answer for the second intercept term (tau2) and
different signs on the coefficients (beta). As long as you use
the same parameterization to calculate the quantities of
interest, the predicted probability of falling into each bin
will be invariant to reparameterization of the systematic
component.
I buy that; what I don't understand is why you gould get different signs
on the coefficients (beta) when you are getting the same estimates (signs
and magnitudes) on the taus. 3) The
substantive question of interest can't be the unobserved
underlying variable (because the stochastic component is
multinomial). Thus, the expected value for this model is the
predicted probability that an observation i falls into each bin.
So let's say that you're interested in presidential approval.
Someone conducts a survey asking: "The president is doing an
effective job. Strongly agree (1), agree (2), neutral (3),
disagree (4), strongly disagree (5)." We observe the number of
people in bins 1:5 and estimate the model and find beta and
tau2-tau5. If we were just interested in the unobserved
underlying distribution Y*, the quantity we would calculate is
x'beta = mu. Now what does mu mean? mu relative to what? It
has to be mu relative to the cut points.
This is what I don't understand. Say you were interested in the effects
of certain covariates on the heights of children. You don't collect the
data, and for some reason the researchers who did collect the data only
tell you which of five categories they are in - really short, kind of
short, average, kind of tall, really tall - and you don't know what the
cutpoints were that they used to divide the children into categories.
Maybe I'm missing something, but this would seem to be the kind of
question one might be interested in answering using something like ordered
probit. In this case, what bin you are in is of no substantive interest,
it is just an artifact of the data collection process. It is the mu that
we are interested in, even if we don't have it in meaningful units. Now,
if I estimate this model with zelig and the coefficient on (say) income is
negative and in polr it is positive, what is one supposed to think?
Cheers
Mike
Olivia
----- Original Message -----
From: "Michael Richard Kellermann" <kellerm(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Thursday, December 16, 2004 10:24 AM
Subject: [gov2001-l] Re: ordered probit coefficients
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
Hi -
I know that we are not supposed to be interested in the raw
coefficient
estimates from something like ordered probit, but how should
we think
about the fact that the coefficient estimates from Zelig are
of the
opposite sign while the intercept/threshold estimates are of
the same sign
as what we are getting from our own code (and from what you
get using
polr() in the MASS package)? What if the substantive question
of interest
is the underlying unobserved variable?
Cheers
Mike
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
4:11 p.m.
New subject: [gov2001-l] mix package
This code should work for the mix package on R 2.0.1:
library(mix)
immig <- read.table("C:/immigration.dat", header = TRUE)
# This turns gender into a 1,2 categorical variable
# We treat the three ordinal variables as continuous rather than
# categorical; otherwise, it has to estimate 686 cells, which is way
# too many given the amount of data.
immig$gender <- immig$gender + 1
# This puts gender on the left of the matrix
immig <- cbind(immig$gender, immig[,1:4])
names(immig) <- c("gender", names(immig[,2:5]))
# This does a whole bunch of stuff to identify the missing data
# I think the data needs to be a matrix, not a frame
s <- prelim.mix(as.matrix(immig), 1)
# This calculates initial estimates of the means and var-covar matrix
thetahat <- em.mix(s)
# This sets the random number seed
rngseed(1234567)
# This generates the imputed dataset. It cranks through the da.mix
# function 5000 times, draws a dataset, and then picks up where it left
# off. A new dataset is drawn after every 5000 steps.
newtheta <- da.mix(s, thetahat, steps = 5000)
getparam.mix(s, newtheta)
x1 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x2 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x3 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x4 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x5 <- as.data.frame(imp.mix(s, newtheta, immig))
# This coerces ipip back into an ordinal variable
x1$ipip <- round(x1$ipip)
x2$ipip <- round(x2$ipip)
x3$ipip <- round(x3$ipip)
x4$ipip <- round(x4$ipip)
x5$ipip <- round(x5$ipip)
# This gets rid of any 6s that creep into the imputed data
# If you end up with any 0s after the round step, get rid of
# those too.
x1$ipip[x1$ipip >5] <- 5
x2$ipip[x2$ipip >5] <- 5
x3$ipip[x3$ipip >5] <- 5
x4$ipip[x4$ipip >5] <- 5
x5$ipip[x5$ipip >5] <- 5
#This turns the imputed datasets into a list for Zelig
immig.mi <- list(x1,x2,x3,x4,x5)
Cheers,
Mike
On Wed, 15 Dec 2004, Olivia Lau wrote:
You should follow the steps in the example at the
bottom of
help(imp.mix)...so yes, you do need to do some pre-processing before you
use imp.mix().
Also, note that I forgot to mention something in section last week: If we
fix tau1 = 0, we need to fix tau2 > 0. So you need to exp(par[k+1]) it to
contstrain it to be greater than zero, but I'm sure that everyone has
figured that out by now... 8)
Olivia.
On Wed, 15 Dec 2004, Jens Hainmueller wrote:
is 'imp.mix' the right command to use
in the mix package for the
imputations? it's the only imputation type command i cound find there.
'imp.mix' apparently requires pre-processing with 'da.mix' or
'dabipf.mix'
do we need to use any of those too?
thanks!
in da mix,
j.
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
4:21 p.m.
New subject: [gov2001-l] mix package
Mike gets, I think, major brownie points for this. 8) This
means that everyone should at least try multiple imputation for
their papers and report whether their coefficients and
quantities of interest display any bias before imputation. 8)
----- Original Message -----
From: "Michael Richard Kellermann" <kellerm(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Thursday, December 16, 2004 9:11 PM
Subject: Re: [gov2001-l] mix package
This code should work for the mix package on R 2.0.1:
library(mix)
immig <- read.table("C:/immigration.dat", header = TRUE)
# This turns gender into a 1,2 categorical variable
# We treat the three ordinal variables as continuous rather
than
# categorical; otherwise, it has to estimate 686 cells, which
is way
# too many given the amount of data.
immig$gender <- immig$gender + 1
# This puts gender on the left of the matrix
immig <- cbind(immig$gender, immig[,1:4])
names(immig) <- c("gender", names(immig[,2:5]))
# This does a whole bunch of stuff to identify the missing
data
# I think the data needs to be a matrix, not a frame
s <- prelim.mix(as.matrix(immig), 1)
# This calculates initial estimates of the means and var-covar
matrix
thetahat <- em.mix(s)
# This sets the random number seed
rngseed(1234567)
# This generates the imputed dataset. It cranks through the
da.mix
# function 5000 times, draws a dataset, and then picks up
where it left
# off. A new dataset is drawn after every 5000 steps.
newtheta <- da.mix(s, thetahat, steps = 5000)
getparam.mix(s, newtheta)
x1 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x2 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x3 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x4 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x5 <- as.data.frame(imp.mix(s, newtheta, immig))
# This coerces ipip back into an ordinal variable
x1$ipip <- round(x1$ipip)
x2$ipip <- round(x2$ipip)
x3$ipip <- round(x3$ipip)
x4$ipip <- round(x4$ipip)
x5$ipip <- round(x5$ipip)
# This gets rid of any 6s that creep into the imputed data
# If you end up with any 0s after the round step, get rid of
# those too.
x1$ipip[x1$ipip >5] <- 5
x2$ipip[x2$ipip >5] <- 5
x3$ipip[x3$ipip >5] <- 5
x4$ipip[x4$ipip >5] <- 5
x5$ipip[x5$ipip >5] <- 5
#This turns the imputed datasets into a list for Zelig
immig.mi <- list(x1,x2,x3,x4,x5)
Cheers,
Mike
On Wed, 15 Dec 2004, Olivia Lau wrote:
You should follow the steps in the example at the
bottom of
help(imp.mix)...so yes, you do need to do some pre-processing
before you
use imp.mix().
Also, note that I forgot to mention something in section last
week: If we
fix tau1 = 0, we need to fix tau2 > 0. So you need to
exp(par[k+1]) it to
contstrain it to be greater than zero, but I'm sure that
everyone has
figured that out by now... 8)
Olivia.
On Wed, 15 Dec 2004, Jens Hainmueller wrote:
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
is 'imp.mix' the right command to use
in the mix package
for the
imputations? it's the only imputation type command i cound
find there.
'imp.mix' apparently requires pre-processing with 'da.mix'
or 'dabipf.mix'
do we need to use any of those too?
thanks!
in da mix,
j.
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
4:46 p.m.
New subject: [gov2001-l] mix package
As a heads up, this was the product of a lot of collective frustration,
error messages, and core dumps on the part of myself, Jens, Ben, Federico,
and others. On other data sets, here are some things to watch out for (some
of which are in the notes in the code) on other datasets:
1) Try to use as few categorical variables as possible; reserve them for
things that really have no logical ordering. If you get an empty cells
error message, this might be the problem.
2) If you declare a variable to be categorical, it really can't have zeros
in it. This causes R to crash on my computer, and I think it is the
reason for the core dumps on the server.
3) If a continuous variable is strongly skewed, try transforming it so
that it is more approximately normal. This can be a reason for the em.mix
stage to go haywire.
4) As always, watch out for linear dependencies in the data. I
accidentally put log(x) and log(x^2) in at the same time and it wasn't
happy. Likewise, if you have a dummy variable (say a country) for which
all of the values of a different variable are missing (say GDP), then you
can't include the dummy; there is nothing to pin down the values.
5) Make sure the data object is a matrix before you start the mix
functions. Leaving it as a data frame produces a list cannot be coerced
to double error message.
6) Most of the problems seem to arise when the data doesn't look exactly
like you think it should look; look at it before running the mix
functions.
Cheers,
Mike
On Thu, 16 Dec 2004, Olivia Lau wrote:
Mike gets, I think, major brownie points for this. 8)
This
means that everyone should at least try multiple imputation for
their papers and report whether their coefficients and
quantities of interest display any bias before imputation. 8)
19 Dec
19 Dec
2:59 p.m.
New subject: [gov2001-l] mix package
Hi Olivia,
I had a question for question no 2. What does Rubin rules mean exactly and
what are we supposed to do? Also, why do we use the mix package in R rather
than the Amelia for this week's problem set?
Thanks!
Wei
Sincerely,
Wei Ha
PhD Candidate in Public Policy
Harvard University
Fax: 1-801-605-1455
----- Original Message -----
From: "Olivia Lau" <olau(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Thursday, December 16, 2004 9:21 PM
Subject: Re: [gov2001-l] mix package
Mike gets, I think, major brownie points for this. 8)
This means that
everyone should at least try multiple imputation for their papers and
report whether their coefficients and quantities of interest display any
bias before imputation. 8)
----- Original Message -----
From: "Michael Richard Kellermann" <kellerm(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Thursday, December 16, 2004 9:11 PM
Subject: Re: [gov2001-l] mix package
This code should work for the mix package on R 2.0.1:
library(mix)
immig <- read.table("C:/immigration.dat", header = TRUE)
# This turns gender into a 1,2 categorical variable
# We treat the three ordinal variables as continuous rather than
# categorical; otherwise, it has to estimate 686 cells, which is way
# too many given the amount of data.
immig$gender <- immig$gender + 1
# This puts gender on the left of the matrix
immig <- cbind(immig$gender, immig[,1:4])
names(immig) <- c("gender", names(immig[,2:5]))
# This does a whole bunch of stuff to identify the missing data
# I think the data needs to be a matrix, not a frame
s <- prelim.mix(as.matrix(immig), 1)
# This calculates initial estimates of the means and var-covar matrix
thetahat <- em.mix(s)
# This sets the random number seed
rngseed(1234567)
# This generates the imputed dataset. It cranks through the da.mix
# function 5000 times, draws a dataset, and then picks up where it left
# off. A new dataset is drawn after every 5000 steps.
newtheta <- da.mix(s, thetahat, steps = 5000)
getparam.mix(s, newtheta)
x1 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x2 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x3 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x4 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x5 <- as.data.frame(imp.mix(s, newtheta, immig))
# This coerces ipip back into an ordinal variable
x1$ipip <- round(x1$ipip)
x2$ipip <- round(x2$ipip)
x3$ipip <- round(x3$ipip)
x4$ipip <- round(x4$ipip)
x5$ipip <- round(x5$ipip)
# This gets rid of any 6s that creep into the imputed data
# If you end up with any 0s after the round step, get rid of
# those too.
x1$ipip[x1$ipip >5] <- 5
x2$ipip[x2$ipip >5] <- 5
x3$ipip[x3$ipip >5] <- 5
x4$ipip[x4$ipip >5] <- 5
x5$ipip[x5$ipip >5] <- 5
#This turns the imputed datasets into a list for Zelig
immig.mi <- list(x1,x2,x3,x4,x5)
Cheers,
Mike
On Wed, 15 Dec 2004, Olivia Lau wrote:
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
You should follow the steps in the example at the
bottom of
help(imp.mix)...so yes, you do need to do some pre-processing before you
use imp.mix().
Also, note that I forgot to mention something in section last week: If
we
fix tau1 = 0, we need to fix tau2 > 0. So you need to exp(par[k+1]) it
to
contstrain it to be greater than zero, but I'm sure that everyone has
figured that out by now... 8)
Olivia.
On Wed, 15 Dec 2004, Jens Hainmueller wrote:
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
is 'imp.mix' the right command to use
in the mix package for the
imputations? it's the only imputation type command i cound find there.
'imp.mix' apparently requires pre-processing with 'da.mix' or
'dabipf.mix'
do we need to use any of those too?
thanks!
in da mix,
j.
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
3:16 p.m.
New subject: [gov2001-l] mix package
Hi Wei,
Please, find the attached files below. It's a Schafer's presentation on MI and
the mix package manual (if you don't have one). Schafer's is a good reading
prior to running the mix package.
Regards,
Paul.
?? weiha <weiha(a)fas.harvard.edu>du>:
Hi Olivia,
I had a question for question no 2. What does Rubin rules mean exactly and
what are we supposed to do? Also, why do we use the mix package in R rather
than the Amelia for this week's problem set?
Thanks!
Wei
Sincerely,
Wei Ha
PhD Candidate in Public Policy
Harvard University
Fax: 1-801-605-1455
----- Original Message -----
From: "Olivia Lau" <olau(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Thursday, December 16, 2004 9:21 PM
Subject: Re: [gov2001-l] mix package
Mike gets, I think, major brownie points for
this. 8) This means that
everyone should at least try multiple imputation for their papers and
report whether their coefficients and quantities of interest display any
bias before imputation. 8)
----- Original Message -----
From: "Michael Richard Kellermann" <kellerm(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Thursday, December 16, 2004 9:11 PM
Subject: Re: [gov2001-l] mix package
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
This code should work for the mix package on R 2.0.1:
library(mix)
immig <- read.table("C:/immigration.dat", header = TRUE)
# This turns gender into a 1,2 categorical variable
# We treat the three ordinal variables as continuous rather than
# categorical; otherwise, it has to estimate 686 cells, which is way
# too many given the amount of data.
immig$gender <- immig$gender + 1
# This puts gender on the left of the matrix
immig <- cbind(immig$gender, immig[,1:4])
names(immig) <- c("gender", names(immig[,2:5]))
# This does a whole bunch of stuff to identify the missing data
# I think the data needs to be a matrix, not a frame
s <- prelim.mix(as.matrix(immig), 1)
# This calculates initial estimates of the means and var-covar matrix
thetahat <- em.mix(s)
# This sets the random number seed
rngseed(1234567)
# This generates the imputed dataset. It cranks through the da.mix
# function 5000 times, draws a dataset, and then picks up where it left
# off. A new dataset is drawn after every 5000 steps.
newtheta <- da.mix(s, thetahat, steps = 5000)
getparam.mix(s, newtheta)
x1 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x2 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x3 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x4 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x5 <- as.data.frame(imp.mix(s, newtheta, immig))
# This coerces ipip back into an ordinal variable
x1$ipip <- round(x1$ipip)
x2$ipip <- round(x2$ipip)
x3$ipip <- round(x3$ipip)
x4$ipip <- round(x4$ipip)
x5$ipip <- round(x5$ipip)
# This gets rid of any 6s that creep into the imputed data
# If you end up with any 0s after the round step, get rid of
# those too.
x1$ipip[x1$ipip >5] <- 5
x2$ipip[x2$ipip >5] <- 5
x3$ipip[x3$ipip >5] <- 5
x4$ipip[x4$ipip >5] <- 5
x5$ipip[x5$ipip >5] <- 5
#This turns the imputed datasets into a list for Zelig
immig.mi <- list(x1,x2,x3,x4,x5)
Cheers,
Mike
On Wed, 15 Dec 2004, Olivia Lau wrote:
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
You should follow the steps in the example at the
bottom of
help(imp.mix)...so yes, you do need to do some pre-processing before you
use imp.mix().
Also, note that I forgot to mention something in section last week: If
we
fix tau1 = 0, we need to fix tau2 > 0. So you need to exp(par[k+1]) it
to
contstrain it to be greater than zero, but I'm sure that everyone has
figured that out by now... 8)
Olivia.
On Wed, 15 Dec 2004, Jens Hainmueller wrote:
> is 'imp.mix' the right command to use in the mix package for the
> imputations? it's the only imputation type command i cound find there.
>
> 'imp.mix' apparently requires pre-processing with 'da.mix' or
> 'dabipf.mix'
> do we need to use any of those too?
>
> thanks!
>
> in da mix,
> j.
>
>
> _______________________________________________
> gov2001-l mailing list
> gov2001-l(a)lists.fas.harvard.edu
> http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
>
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
3:27 p.m.
New subject: [gov2001-l] mix package
Hi, Wei.
The rubin rules are described in the rubin reading that gary
assigned in the lecture notes. In practice, it means: "Use
Zelig because it does it for you." 8)
We are using mix because it's in R. Amelia isn't in R (yet).
Olivia
----- Original Message -----
From: "weiha" <weiha(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Sunday, December 19, 2004 7:58 PM
Subject: Re: [gov2001-l] mix package
Hi Olivia,
I had a question for question no 2. What does Rubin rules mean
exactly and what are we supposed to do? Also, why do we use
the mix package in R rather than the Amelia for this week's
problem set?
Thanks!
Wei
Sincerely,
Wei Ha
PhD Candidate in Public Policy
Harvard University
Fax: 1-801-605-1455
----- Original Message -----
From: "Olivia Lau" <olau(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Thursday, December 16, 2004 9:21 PM
Subject: Re: [gov2001-l] mix package
Mike gets, I think, major brownie points for
this. 8) This
means that everyone should at least try multiple imputation
for their papers and report whether their coefficients and
quantities of interest display any bias before imputation.
8)
----- Original Message -----
From: "Michael Richard Kellermann" <kellerm(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Thursday, December 16, 2004 9:11 PM
Subject: Re: [gov2001-l] mix package
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
This code should work for the mix package on R 2.0.1:
library(mix)
immig <- read.table("C:/immigration.dat", header = TRUE)
# This turns gender into a 1,2 categorical variable
# We treat the three ordinal variables as continuous rather
than
# categorical; otherwise, it has to estimate 686 cells,
which is way
# too many given the amount of data.
immig$gender <- immig$gender + 1
# This puts gender on the left of the matrix
immig <- cbind(immig$gender, immig[,1:4])
names(immig) <- c("gender", names(immig[,2:5]))
# This does a whole bunch of stuff to identify the missing
data
# I think the data needs to be a matrix, not a frame
s <- prelim.mix(as.matrix(immig), 1)
# This calculates initial estimates of the means and
var-covar matrix
thetahat <- em.mix(s)
# This sets the random number seed
rngseed(1234567)
# This generates the imputed dataset. It cranks through the
da.mix
# function 5000 times, draws a dataset, and then picks up
where it left
# off. A new dataset is drawn after every 5000 steps.
newtheta <- da.mix(s, thetahat, steps = 5000)
getparam.mix(s, newtheta)
x1 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x2 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x3 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x4 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x5 <- as.data.frame(imp.mix(s, newtheta, immig))
# This coerces ipip back into an ordinal variable
x1$ipip <- round(x1$ipip)
x2$ipip <- round(x2$ipip)
x3$ipip <- round(x3$ipip)
x4$ipip <- round(x4$ipip)
x5$ipip <- round(x5$ipip)
# This gets rid of any 6s that creep into the imputed data
# If you end up with any 0s after the round step, get rid of
# those too.
x1$ipip[x1$ipip >5] <- 5
x2$ipip[x2$ipip >5] <- 5
x3$ipip[x3$ipip >5] <- 5
x4$ipip[x4$ipip >5] <- 5
x5$ipip[x5$ipip >5] <- 5
#This turns the imputed datasets into a list for Zelig
immig.mi <- list(x1,x2,x3,x4,x5)
Cheers,
Mike
On Wed, 15 Dec 2004, Olivia Lau wrote:
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
You should follow the steps in the example at the
bottom of
help(imp.mix)...so yes, you do need to do some
pre-processing before you
use imp.mix().
Also, note that I forgot to mention something in section
last week: If we
fix tau1 = 0, we need to fix tau2 > 0. So you need to
exp(par[k+1]) it to
contstrain it to be greater than zero, but I'm sure that
everyone has
figured that out by now... 8)
Olivia.
On Wed, 15 Dec 2004, Jens Hainmueller wrote:
> is 'imp.mix' the right command to use in the mix package
> for the
> imputations? it's the only imputation type command i
> cound find there.
>
> 'imp.mix' apparently requires pre-processing with
> 'da.mix' or 'dabipf.mix'
> do we need to use any of those too?
>
> thanks!
>
> in da mix,
> j.
>
>
> _______________________________________________
> gov2001-l mailing list
> gov2001-l(a)lists.fas.harvard.edu
> http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
>
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
3:30 p.m.
New subject: [gov2001-l] mix package
you're welcome to use Amelia instead of course, but not in R.
Gary
On Sun, 19 Dec 2004, Olivia Lau wrote:
Hi, Wei.
The rubin rules are described in the rubin reading that gary
assigned in the lecture notes. In practice, it means: "Use
Zelig because it does it for you." 8)
We are using mix because it's in R. Amelia isn't in R (yet).
Olivia
----- Original Message -----
From: "weiha" <weiha(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Sunday, December 19, 2004 7:58 PM
Subject: Re: [gov2001-l] mix package
Hi Olivia,
I had a question for question no 2. What does Rubin rules mean
exactly and what are we supposed to do? Also, why do we use
the mix package in R rather than the Amelia for this week's
problem set?
Thanks!
Wei
Sincerely,
Wei Ha
PhD Candidate in Public Policy
Harvard University
Fax: 1-801-605-1455
----- Original Message -----
From: "Olivia Lau" <olau(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Thursday, December 16, 2004 9:21 PM
Subject: Re: [gov2001-l] mix package
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
Mike gets, I think, major brownie points for
this. 8) This
means that everyone should at least try multiple imputation
for their papers and report whether their coefficients and
quantities of interest display any bias before imputation.
8)
----- Original Message -----
From: "Michael Richard Kellermann" <kellerm(a)fas.harvard.edu>
To: <gov2001-l(a)lists.fas.harvard.edu>
Sent: Thursday, December 16, 2004 9:11 PM
Subject: Re: [gov2001-l] mix package
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
This code should work for the mix package on R 2.0.1:
library(mix)
immig <- read.table("C:/immigration.dat", header = TRUE)
# This turns gender into a 1,2 categorical variable
# We treat the three ordinal variables as continuous rather
than
# categorical; otherwise, it has to estimate 686 cells,
which is way
# too many given the amount of data.
immig$gender <- immig$gender + 1
# This puts gender on the left of the matrix
immig <- cbind(immig$gender, immig[,1:4])
names(immig) <- c("gender", names(immig[,2:5]))
# This does a whole bunch of stuff to identify the missing
data
# I think the data needs to be a matrix, not a frame
s <- prelim.mix(as.matrix(immig), 1)
# This calculates initial estimates of the means and
var-covar matrix
thetahat <- em.mix(s)
# This sets the random number seed
rngseed(1234567)
# This generates the imputed dataset. It cranks through the
da.mix
# function 5000 times, draws a dataset, and then picks up
where it left
# off. A new dataset is drawn after every 5000 steps.
newtheta <- da.mix(s, thetahat, steps = 5000)
getparam.mix(s, newtheta)
x1 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x2 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x3 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x4 <- as.data.frame(imp.mix(s, newtheta, immig))
newtheta <- da.mix(s, newtheta, steps = 5000)
x5 <- as.data.frame(imp.mix(s, newtheta, immig))
# This coerces ipip back into an ordinal variable
x1$ipip <- round(x1$ipip)
x2$ipip <- round(x2$ipip)
x3$ipip <- round(x3$ipip)
x4$ipip <- round(x4$ipip)
x5$ipip <- round(x5$ipip)
# This gets rid of any 6s that creep into the imputed data
# If you end up with any 0s after the round step, get rid of
# those too.
x1$ipip[x1$ipip >5] <- 5
x2$ipip[x2$ipip >5] <- 5
x3$ipip[x3$ipip >5] <- 5
x4$ipip[x4$ipip >5] <- 5
x5$ipip[x5$ipip >5] <- 5
#This turns the imputed datasets into a list for Zelig
immig.mi <- list(x1,x2,x3,x4,x5)
Cheers,
Mike
On Wed, 15 Dec 2004, Olivia Lau wrote:
> You should follow the steps in the example at the bottom of
> help(imp.mix)...so yes, you do need to do some
> pre-processing before you
> use imp.mix().
>
> Also, note that I forgot to mention something in section
> last week: If we
> fix tau1 = 0, we need to fix tau2 > 0. So you need to
> exp(par[k+1]) it to
> contstrain it to be greater than zero, but I'm sure that
> everyone has
> figured that out by now... 8)
>
> Olivia.
>
> On Wed, 15 Dec 2004, Jens Hainmueller wrote:
>
> > is 'imp.mix' the right command to use in the mix package
> > for the
> > imputations? it's the only imputation type command i
> > cound find there.
> >
> > 'imp.mix' apparently requires pre-processing with
> > 'da.mix' or 'dabipf.mix'
> > do we need to use any of those too?
> >
> > thanks!
> >
> > in da mix,
> > j.
> >
> >
> > _______________________________________________
> > gov2001-l mailing list
> > gov2001-l(a)lists.fas.harvard.edu
> > http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
> >
> _______________________________________________
> gov2001-l mailing list
> gov2001-l(a)lists.fas.harvard.edu
> http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
>
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
20 Dec
20 Dec
7:01 a.m.
New subject: [gov2001-l] assignment 9
Olivia,
I attached a pdf copy of assignment 9. Happy holidays,
Federico
7 Feb
7 Feb
8:23 p.m.
New subject: [gov2001-l] paper
Hey Mike,
How's the semester and generals studying going? Busy here, but regardless,
I was wondering if you would be interested in taking the civil wars paper a
bit further and if all goes well submitting it to some journals. Gary's
comments on the back seem very optimistic, and from talking to Olivia, he
apparently liked our paper best in the class. Let me know your thoughts.
Best,
Amit
PS~ I actually can't find an R code file for the paper, mind sending me it?
8:40 p.m.
New subject: [gov2001-l] paper
Sorry for awakening a finally restful list, my unfounded hearsay, and the
spam from accidentally replying to the wrong address.
Apologetically,
Amit
7036
days inactive
7097
days old
gov2001@lists.gking.harvard.edu
17 comments
8 participants
participants (8)
-
ferrara＠fas.harvard.edu -
jens_hainmueller＠ksg05.harvard.edu -
kellerm＠fas.harvard.edu -
king＠harvard.edu -
modi＠fas.harvard.edu -
mschoi＠fas.harvard.edu -
olau＠fas.harvard.edu -
weiha＠fas.harvard.edu