R help - Dec 2010 - Specifying Prior Weights in a GLM

Hello R folks,

I have three questions. I am trying to run a logistic regression (binomial
family) where the response variable is a proportion.  According to R
Documentation in "a binomial GLM prior weights are used to give the number
of trials when the response is the proportion of successes."  However when
I
run my code I get the following error message:

Error in model.frame.default(formula = PER_ELA ~ A_EX + COMM + ENG + S_R + 
: 
  variable lengths differ (found for '(weights)')

I'm not sure what I am doing wrong.  My response variable is Y/M, which is
the proportion of 1's (successes) among M binary responses.  My prior weight
is a variable indicating the number of trials for each observation.  


This is an abbreviated version of the code that I ran: 

glm1<-glm(PER_ELA~A_EX .... PER_LEA,
family=binomial(link="logit"),data=data2,weights="REG")


Question 1 and 2:
Does the number of trials for each observation in my dataset have to be the
same? What am I doing wrong here?


Question 3:
Is it OK for me to use percentages as predictor variables in a logistic
regression? 

-- 
View this message in context:
http://r.789695.n4.nabble.com/Specifying-Prior-Weights-in-a-GLM-tp3083480p3083480.html
Sent from the R help mailing list archive at Nabble.com.

Hi:

Here's a simple example of its use using the data from Montgomery, Myers and
Vining (2002), problem 4.7:

ex4.7 <- structure(list(pressure = c(2500L, 2700L, 2900L, 3100L, 3300L,
3500L, 3700L, 3900L, 4100L, 4300L), nfast = c(50L, 70L, 100L,
60L, 40L, 85L, 90L, 50L, 80L, 65L), nfailures = c(10L, 17L, 30L,
21L, 18L, 43L, 54L, 33L, 60L, 51L)), .Names = c("pressure",
"nfast",
"nfailures"), row.names = c("1", "2",
"3", "4", "5", "6", "7",
"8", "9", "10"), class = "data.frame")

# nfast is the sample size per covariate class, nfailures and pressure
# the response and covariate, respectively.

# Notice that the weights variable is the denominator of the response
m <- glm(nfailures/nfast ~ pressure, data = ex4.7, weights = nfast)

ex4.7a <- cbind(ex4.7, pred = predict(m, type = 'response'))

# Check that the observed proportions and the predicted probabilities are
reasonably in sync

library(ggplot2)
ggplot(ex4.7a, aes(x = pressure, y = nfailures/nfast)) +
    geom_point(size = 2.5) + geom_line(aes(y = pred), size = 1, color
'red')

# or in base graphics,
plot(nfailures/nfast ~ pressure, data = ex4.7a, pch = 16)
with(ex4.7a, lines(pressure, pred, col = 'red', lwd = 1.2))

HTH,
Dennis

On Sat, Dec 11, 2010 at 8:59 AM, Enrique Garcia <eckomorph@yahoo.com>
wrote:
>
> Hello R folks,
>
> I have three questions. I am trying to run a logistic regression (binomial
> family) where the response variable is a proportion.  According to R
> Documentation in "a binomial GLM prior weights are used to give the
number
> of trials when the response is the proportion of successes."  However
when
> I
> run my code I get the following error message:
>
> Error in model.frame.default(formula = PER_ELA ~ A_EX + COMM + ENG + S_R +
> :
>  variable lengths differ (found for '(weights)')
>
> I'm not sure what I am doing wrong.  My response variable is Y/M, which
is
> the proportion of 1's (successes) among M binary responses.  My prior
> weight
> is a variable indicating the number of trials for each observation.
>
>
> This is an abbreviated version of the code that I ran:
>
> glm1<-glm(PER_ELA~A_EX .... PER_LEA,
> family=binomial(link="logit"),data=data2,weights="REG")
>
>
> Question 1 and 2:
> Does the number of trials for each observation in my dataset have to be the
> same? What am I doing wrong here?
>
>
> Question 3:
> Is it OK for me to use percentages as predictor variables in a logistic
> regression?
>
> --
> View this message in context:
>
http://r.789695.n4.nabble.com/Specifying-Prior-Weights-in-a-GLM-tp3083480p3083480.html
> Sent from the R help mailing list archive at Nabble.com.
>
> ______________________________________________
> R-help@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.
>
	[[alternative HTML version deleted]]

Hey Dennis,

Thank you.  This was helpful.  It looks like the structure of my code was
incorrect.  I was using a variable containing proportions as my D.V. instead
of using a two column vector.  I re-ran the code using the following: 

Y<-(successes/trials)

-Enrique 
-- 
View this message in context:
http://r.789695.n4.nabble.com/Specifying-Prior-Weights-in-a-GLM-tp3083480p3084674.html
Sent from the R help mailing list archive at Nabble.com.