R help - Jan 2011 - Logical Indicator for Warning and Error Messages?

Dear R Community-

 

Is there a logical variable indicating the presence of a warning or error
message after executing a command?  I’m bootstrapping a logit model (1,000
iterations, relevant code posted below), but the model fails to converge on
many iterations.  (I’m guessing that a small sample size and the combination
of sampling with replacement, high item missingness, and list-wise deletion
leads to a sparse data problem for some iterations.)    

 

Even for the iterations that fail to converge, though, the program produces
parameter estimates.   I would like to remove these parameter estimates from
N x k matrix containing the boostrapped parameter estimates.  What I was
thinking of is adding a line of code to check if there is a warning or error
message present, recording the iteration numbers where there are
error/warning messages, and replacing the parameters for those iterations
with NA’s.  I’ve checked the help pages for the “warning”, “getOption”, and
other commands, but it’s not immediately apparent to me how to check for the
presence of warning or error messages.  

 

Here’s the relevant portion of the code (I’ve omitted the parts that recover
parameter and fit estimates for each iteration):

 

*************************************

[Code Starts Here]



#Create Objects

N = 1000                               #1000 iterations

set.seed(7843)

seed = sample(10000000, 1000)



#1000 Iterations

for (i in 1:N){



#Bootstap (Sampling with Replacement)

set.seed(seed[i])

pre <- subset(cainit, time==1)

sampt1 <- sort(sample(nrow(pre), nrow(pre), replace=T))

sampt1 <- pre[sampt1,]





#T1 Model of Vote Intention

try(T1A <- glm(prop1a ~ educ + age + inc + dem + appgov + appleg + budget +
co1a + 

                prop1b + prop1d,family=binomial(link="logit"), 

                data=sampt1, na.action=na.exclude))



##[NOTE:  the “lrm” command is part of the “Design” package]

try(T1a <- lrm(prop1a ~ educ + age + inc + dem + appgov + appleg + budget +
co1a + 

                prop1b + prop1d, data=sampt1))

}



[Code Ends Here]

******************************************







The following set of warning/error messages is repeated about a dozen times
at the end of the loop:



Error in lrm(prop1a ~ educ + age + inc + dem + appgov + appleg + budget +  :


  Unable to fit model using “lrm.fit”

In addition: Warning messages:

1: glm.fit: fitted probabilities numerically 0 or 1 occurred 

2: glm.fit: fitted probabilities numerically 0 or 1 occurred 

3: glm.fit: algorithm did not converge 

4: glm.fit: fitted probabilities numerically 0 or 1 occurred







Any help would be greatly appreciated.  



Happy New Year!

David



=============================
David Crow

Profesor-Investigador / Assistant Professor

División de Estudios Internacionales

Centro de Investigación y Docencia Económicas


Carretera México-Toluca 3655
Col. Lomas de Santa Fe
01210 México, D.F.
Tel.:   (+011 52 55) 5727-9800, ext. 2152



=============================



	[[alternative HTML version deleted]]

On Jan 2, 2011, at 11:41 PM, David Crow wrote:
> Dear R Community-
>
>
>
> Is there a logical variable indicating the presence of a warning or  
> error
> message after executing a command?  I?m bootstrapping a logit model  
> (1,000
> iterations, relevant code posted below), but the model fails to  
> converge on
> many iterations.  (I?m guessing that a small sample size and the  
> combination
> of sampling with replacement, high item missingness, and list-wise  
> deletion
> leads to a sparse data problem for some iterations.)
>
>
>
> Even for the iterations that fail to converge, though, the program  
> produces
> parameter estimates.   I would like to remove these parameter  
> estimates from
> N x k matrix containing the boostrapped parameter estimates.  What I  
> was
> thinking of is adding a line of code to check if there is a warning  
> or error
> message present, recording the iteration numbers where there are
> error/warning messages, and replacing the parameters for those  
> iterations
> with NA?s.  I?ve checked the help pages for the ?warning?,  
> ?getOption?, and
> other commands, but it?s not immediately apparent to me how to check  
> for the
> presence of warning or error messages.
?options
#says that if "warn" is >= 2 that warnings are turned into errors
?try
# for handling of errors

-- 
David.
>
>
>
> Here?s the relevant portion of the code (I?ve omitted the parts that  
> recover
> parameter and fit estimates for each iteration):
>
>
>
> *************************************
>
> [Code Starts Here]
>
>
>
> #Create Objects
>
> N = 1000                               #1000 iterations
>
> set.seed(7843)
>
> seed = sample(10000000, 1000)
>
>
>
> #1000 Iterations
>
> for (i in 1:N){
>
>
>
> #Bootstap (Sampling with Replacement)
>
> set.seed(seed[i])
>
> pre <- subset(cainit, time==1)
>
> sampt1 <- sort(sample(nrow(pre), nrow(pre), replace=T))
>
> sampt1 <- pre[sampt1,]
>
>
>
>
>
> #T1 Model of Vote Intention
>
> try(T1A <- glm(prop1a ~ educ + age + inc + dem + appgov + appleg +  
> budget +
> co1a +
>
>                prop1b + prop1d,family=binomial(link="logit"),
>
>                data=sampt1, na.action=na.exclude))
>
>
>
> ##[NOTE:  the ?lrm? command is part of the ?Design? package]
>
> try(T1a <- lrm(prop1a ~ educ + age + inc + dem + appgov + appleg +  
> budget +
> co1a +
>
>                prop1b + prop1d, data=sampt1))
>
> }
>
>
>
> [Code Ends Here]
>
> ******************************************
>
>
>
>
>
>
>
> The following set of warning/error messages is repeated about a  
> dozen times
> at the end of the loop:
>
>
>
> Error in lrm(prop1a ~ educ + age + inc + dem + appgov + appleg +  
> budget +  :
>
>
>  Unable to fit model using ?lrm.fit?
>
> In addition: Warning messages:
>
> 1: glm.fit: fitted probabilities numerically 0 or 1 occurred
>
> 2: glm.fit: fitted probabilities numerically 0 or 1 occurred
>
> 3: glm.fit: algorithm did not converge
>
> 4: glm.fit: fitted probabilities numerically 0 or 1 occurred
>
>
>
>
>
>
>
> Any help would be greatly appreciated.
>
>
>
> Happy New Year!
>
> David
>
>
>
> =============================>
> David Crow
>
> Profesor-Investigador / Assistant Professor
>
> Divisi?n de Estudios Internacionales
>
> Centro de Investigaci?n y Docencia Econ?micas
>
>
> Carretera M?xico-Toluca 3655
> Col. Lomas de Santa Fe
> 01210 M?xico, D.F.
> Tel.:   (+011 52 55) 5727-9800, ext. 2152
>
>
>
> =============================>
>
>
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
David Winsemius, MD
West Hartford, CT