5 May
2006
5 May
'06
9:49 a.m.
When trying to impute a binary variable using mice, I encounter this
error message in the first iteration,
imp <- mice(Womendata, imputationMethod=imp.method)
iter imp variable
1 1 Approval.member Approval.Congress Trust.Index
Efficacy.Index Efficacy.Index2 CompetenceError in x[good, ] * w :
non-numeric argument to binary operator
My data table has the following dimension 15502 by 30. Competence is
a binary numeric variable with about 150 NA's. In my imp.method
vector, I have specified "logreg" as the method of imputation. Does
anyone have any suggestions as to what this error could mean.
Regards,
Sheldon
5 May
5 May
10:13 a.m.
Sheldon:
I couldn't seem to get mice running on my mac and kept getting weird
errors. I was working with Elena and, rather than investigate, we
tried the same code on a PC and it worked fine. Before you devote too
much time to debugging your code, you might just see if the same
lines work with the Windows version of R (assuming that's an easy
test). To be clear, our problem may have had nothing to do with being
on a mac but I thought I'd pass along our experience.
Omar
On May 5, 2006, at 1:49 PM, Sheldon Bond wrote:
When trying to impute a binary variable using mice, I encounter this
error message in the first iteration,
imp <- mice(Womendata, imputationMethod=imp.method)
iter imp variable
1 1 Approval.member Approval.Congress Trust.Index
Efficacy.Index Efficacy.Index2 CompetenceError in x[good, ] * w :
non-numeric argument to binary operator
My data table has the following dimension 15502 by 30. Competence is
a binary numeric variable with about 150 NA's. In my imp.method
vector, I have specified "logreg" as the method of imputation. Does
anyone have any suggestions as to what this error could mean.
Regards,
Sheldon
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
Omar Wasow
M.A.-Ph.D. Candidate
Department of African and African American Studies
Department of Government
Harvard University
10:23 a.m.
Good idea. I will give this a try. Oddly enough, I can get it to
work if I adjust the order of the variables.
Regards,
Sheldon
On 5/5/06, Omar Wasow <owasow(a)fas.harvard.edu> wrote:
Sheldon:
I couldn't seem to get mice running on my mac and kept getting weird
errors. I was working with Elena and, rather than investigate, we
tried the same code on a PC and it worked fine. Before you devote too
much time to debugging your code, you might just see if the same
lines work with the Windows version of R (assuming that's an easy
test). To be clear, our problem may have had nothing to do with being
on a mac but I thought I'd pass along our experience.
Omar
On May 5, 2006, at 1:49 PM, Sheldon Bond wrote:
When trying to impute a binary variable using mice, I encounter this
error message in the first iteration,
imp <- mice(Womendata,
imputationMethod=imp.method)
iter imp variable
1 1 Approval.member Approval.Congress Trust.Index
Efficacy.Index Efficacy.Index2 CompetenceError in x[good, ] * w :
non-numeric argument to binary operator
My data table has the following dimension 15502 by 30. Competence is
a binary numeric variable with about 150 NA's. In my imp.method
vector, I have specified "logreg" as the method of imputation. Does
anyone have any suggestions as to what this error could mean.
Regards,
Sheldon
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
Omar Wasow
M.A.-Ph.D. Candidate
Department of African and African American Studies
Department of Government
Harvard University
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
10:55 a.m.
New subject: [gov2001-l] Claire And Janet's Abstract
Here is our abstract. Thanks in advance for any comments.
Claire and Janet
-------------------------------
Why Do Civil Wars Recur?
Living Conditions, Attributes of the Prior War, and Autocracies Matter
Of civil wars that have ended since 1945, roughly one-third have
reignited. Walter (2004) argues that conditions conducive to rebel
recruitment - particularly poor living conditions and lack of access to
political participation - are the critical precursors to renewed civil
war. Using rare-events logit, model specification tests, and a hazard
model on a respecified dataset, we demonstrate that Walter's findings are
robust with three key exceptions: (1) living conditions matter, but less
so than the characteristics of the previous war; (2) the magnitudes of her
variables' impact are smaller than she estimated; and (3) when controlling
for per capita GDP, civil war recurrence is less likely in autocratic
states.
11:29 a.m.
New subject: [gov2001-l] Abstract: Lucy Barnes
There is no pathway by which income inequality affects health (but
somehow it does?)
There is a seemingly robust negative association between the level of
income inequality of a population and its health outcomes. Yet there has
been little investigation of any of the theorized mechanisms for why
this is the case. Using hierarchical models of individuals nested within
states, I find no evidence that income inequality affects health due to
relative material deprivation, nor adverse psycho-social effects on the
badly-off. Nor do I find support for the contention that the social
disintegration caused by inequality mediates the association. I find no
evidence either that racial composition has effects on health via social
cohesion. This places the burden of proof on the shoulders of the
defenders of the `income inequality hypothesis' to demonstrate causation
or consider other channels by which inequality might make its effect, to
counter the charge that the association is spurious.
12:29 p.m.
New subject: [gov2001-l] Abstract: Lucy Barnes
Sounds like it will be a neat paper--and generally, the core message of
the abstract seems quite clear. The title echoes that message and is
quite clear as well but perhaps a bit too colloquial. Other than that, my
main recommendation is to discuss the dataset a bit, so that people know
the population under discussion. From reaidng the below, people might not
know if you are talking about countries, US cities, states (the truth,
right?), counties, etc. The abstract is a littany of failed potential
mechanisms, and that helps make the point, but you might summarize the
hypotheses a bit more quickly in order to talk a bit more about the data
(and perhaps methods) that are allowing you to draw these conclusions.
Best,
Dan
On Fri, 5 May 2006, Lucy Barnes wrote:
There is no pathway by which income inequality affects
health (but
somehow it does?)
There is a seemingly robust negative association between the level of
income inequality of a population and its health outcomes. Yet there has
been little investigation of any of the theorized mechanisms for why
this is the case. Using hierarchical models of individuals nested within
states, I find no evidence that income inequality affects health due to
relative material deprivation, nor adverse psycho-social effects on the
badly-off. Nor do I find support for the contention that the social
disintegration caused by inequality mediates the association. I find no
evidence either that racial composition has effects on health via social
cohesion. This places the burden of proof on the shoulders of the
defenders of the `income inequality hypothesis' to demonstrate causation
or consider other channels by which inequality might make its effect, to
counter the charge that the association is spurious.
_______________________________________________
gov2001-l mailing list
gov2001-l(a)lists.fas.harvard.edu
http://lists.fas.harvard.edu/mailman/listinfo/gov2001-l
2:21 p.m.
New subject: [gov2001-l] Abstract: Lucy Barnes
this one I like more than Dan. the colloquial part adds a little
attitude, but it seems appropriate.
(i'm assuming you got to this abstract by first doing a replication, but
its good that you found a way to generalize it.)
Gary
On Fri, 5 May 2006, Dan Hopkins wrote:
Sounds like it will be a neat paper--and generally,
the core message of
the abstract seems quite clear. The title echoes that message and is
quite clear as well but perhaps a bit too colloquial. Other than that, my
main recommendation is to discuss the dataset a bit, so that people know
the population under discussion. From reaidng the below, people might not
know if you are talking about countries, US cities, states (the truth,
right?), counties, etc. The abstract is a littany of failed potential
mechanisms, and that helps make the point, but you might summarize the
hypotheses a bit more quickly in order to talk a bit more about the data
(and perhaps methods) that are allowing you to draw these conclusions.
Best,
Dan
On Fri, 5 May 2006, Lucy Barnes wrote:
There is no pathway by which income inequality
affects health (but
somehow it does?)
There is a seemingly robust negative association between the level of
income inequality of a population and its health outcomes. Yet there has
been little investigation of any of the theorized mechanisms for why
this is the case. Using hierarchical models of individuals nested within
states, I find no evidence that income inequality affects health due to
relative material deprivation, nor adverse psycho-social effects on the
badly-off. Nor do I find support for the contention that the social
disintegration caused by inequality mediates the association. I find no
evidence either that racial composition has effects on health via social
cohesion. This places the burden of proof on the shoulders of the
defenders of the `income inequality hypothesis' to demonstrate causation
or consider other channels by which inequality might make its effect, to
counter the charge that the association is spurious.
_______________________________________________
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
6578
days inactive
6578
days old
gov2001@lists.gking.harvard.edu
6 comments
6 participants
participants (6)
-
cschwart＠fas.harvard.edu -
dhopkins＠fas.harvard.edu -
king＠harvard.edu -
lbarnes＠fas.harvard.edu -
owasow＠fas.harvard.edu -
sheldon.bond＠gmail.com