Skip to content

controlling number of decimals printed in anova tables?

7 messages · Dieter Menne, Gabor Grothendieck, Michael Friendly

Michael Friendly
# 
For glm() models, I often find both the print() and summary() method 
disappointing if my main interest
is seeing how well a given model fits. A basic display would just 
compare the null model to to my model.

I wrote the function below based on code in some package.
How can I make it so that the df in the table are printed with 0 decimals?

# display basic model fit statistics for a glm object

modelFit.glm <-
function (x, digits = max(3, getOption("digits") - 3), ...)
{
dev <- c(x$null.deviance, x$deviance )
df <- c(x$df.null, x$df.residual)
table <- data.frame(dev, df, c(NA, 1-pchisq(x$deviance, x$df.residual)),
row.names=c("Null model", "Model"))

dimnames(table) <- list(c("Null model", "Model"), c("Deviance", "df", 
"Pr(>Chi^2)"))
title <- paste("Analysis of Deviance Table", "\n\tFormula: ", 
deparse(x$formula), "\n")
structure(table, heading = title, class = c("anova", "data.frame"))

}

berkeley <- as.data.frame(UCBAdmissions)
berk.mod2 <- glm(Freq ~ Dept * (Gender+Admit), data=berkeley, 
family="poisson")

 > modelFit.glm(berk.mod2)
Analysis of Deviance Table
Formula: Freq ~ Dept * (Gender + Admit)

Deviance df Pr(>Chi^2)
Null model 2650.10 23.00
Model 21.74 6.00 0.001352 **
---
Signif. codes: 0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1
 >
Gabor Grothendieck
# 
Try this:
On Mon, Feb 16, 2009 at 9:02 AM, Michael Friendly <friendly at yorku.ca> wrote:
Try this:
df <- as.integer(c(x$df.null, x$df.residual))
Gabor Grothendieck
# 
Or safer:

df <- as.integer(round(...))

On Mon, Feb 16, 2009 at 10:54 AM, Gabor Grothendieck
<ggrothendieck at gmail.com> wrote:
Dieter Menne
# 
Gabor Grothendieck <ggrothendieck <at> gmail.com> writes:
Did you try? I believe it is a problem of printCoefmat that has quite
a few options for special column, but none for df. Ask Martin M?chler.

Dieter
Gabor Grothendieck
# 
On Mon, Feb 16, 2009 at 11:08 AM, Dieter Menne
<dieter.menne at menne-biomed.de> wrote:
Yes, with as.integer(round(...)) It looks like this:
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
Analysis of Deviance Table
        Formula:  Freq ~ Dept * (Gender + Admit)

           Deviance   df Pr(>Chi^2)
Null model     2650   23
Model            22    6     0.0014 **
---
Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Also note:
[1] 0
[1] 1
Dieter Menne
# 
Gabor Grothendieck <ggrothendieck <at> gmail.com> writes:
I am 90% I am on the wrong trip, help me out. And what happens to your solutions
if the Null Model has small Deviance?

Dieter


modelFit.glm <-
function (x, digits = max(3, getOption("digits") - 3), ...)
{
dev <- c(x$null.deviance, x$deviance )
df <- as.integer(c(x$df.null, x$df.residual))
#df <- as.integer(round(c(x$df.null, x$df.residual)))
table <- data.frame(dev, df, c(NA, 1-pchisq(x$deviance, x$df.residual)),
row.names=c("Null model", "Model"))

dimnames(table) <- list(c("Null model", "Model"), c("Deviance", "df", 
"Pr(>Chi^2)"))
title <- paste("Analysis of Deviance Table", "\n\tFormula: ", 
deparse(x$formula), "\n")
structure(table, heading = title, class = c("anova", "data.frame"))

}

berkeley <- as.data.frame(UCBAdmissions)
berk.mod2 <- glm(Freq ~ Dept * (Gender+Admit), data=berkeley, 
family="poisson")

modelFit.glm(berk.mod2)



Analysis of Deviance Table 
        Formula:  Freq ~ Dept * (Gender + Admit) 

           Deviance      df Pr(>Chi^2)   
Null model  2650.10   23.00              
Model         21.74    6.00   0.001352 **
---
Signif. codes:  0 ?***? 0.001 ?**? 0.01 ?*? 0.05 ?.? 0.1 ? ? 1
Michael Friendly
# 
Thanks, Gabor
No, that wasn't it at all.  In print.anova, I found:

    if (length(i <- grep("Df$", cn)))
        zap.i <- zap.i[!(zap.i %in% i)]
 
so it only recognizes "Df", not "df" as a column name prefix to print as 
integers.
-Michael
Gabor Grothendieck wrote: