[R] Problem with bargraph.CI in Sciplot package

[Manuel Morales]

On Fri, 2009-04-10 at 15:07 -0700, Metconnection wrote:
> Hi there, 
> I wonder if anyone can help me. I'm trying to use bargraph.CI in the Sciplot
> package when there is a missing combination of the factor levels.
> Unfortunately the standard errors on the plot do not appear to be correct. 
> 
> Consider an analysis consisting of two factors A and B. When all factor
> level combinations are present all appears fine:
> 
> library(sciplot)
> 
> #all data
> response<-c(32,54,32,65,34,65,65,45,54,23,43,23,76,87,65,45)
> factorA<-c("A","A","A","A","A","A","A","A","B","B","B","B","B","B","B","B")
> factorB<-c("a","a","a","a","b","b","b","b","a","a","a","a","b","b","b","b")
> data<-data.frame(response,factorA,factorB)
> bargraph.CI(x.factor = data$factorB,data$response, group=data$factorA,
> legend=TRUE)
> 
> 
> 
> but when I remove a combination of the two factors the error whiskers appear
> to be on the wrong bars:
> 
> 
> #excluding one combination
> response<-c(34,65,65,45,54,23,43,23,76,87,65,45)
> factorA<-c("A","A","A","A","B","B","B","B","B","B","B","B")
> factorB<-c("b","b","b","b","a","a","a","a","b","b","b","b")
> data<-data.frame(response,factorA,factorB)
> bargraph.CI(x.factor = data$factorB,data$response, group=data$factorA,
> legend=TRUE)
> 
> 
> I had a look at the ci.fun options but with no success.
> 
> Thanks in advance for any ideas
> Simon

Hi Simon,

bargraph.CI was intended for factorial designs and it doesn't work well
with missing factor combinations. I'll take a look at changing this, but
for now, a quick fix is to add the missing combination with an NA value
for the response. Eg., change the data in your second example to:

response<-c(34,65,65,45,54,23,43,23,76,87,65,45,NA)
factorA<-c("A","A","A","A","B","B","B","B","B","B","B","B","A")
factorB<-c("b","b","b","b","a","a","a","a","b","b","b","b","a")


-- 
http://mutualism.williams.edu