hnp (version 1.2-7)

corn: Corn damage data

Description

Corn grain damage by the maize weevil, a major pest of stored maize worldwide.

Usage

data(corn)

Format

A data frame with 40 observations on the following 3 variables.

extract	factor	with levels leaf, branch, seed and control
m	numeric	total number of corn grains
y	numeric	number of damaged corn grains

Details

A major pest of stored maize in Brazil is Sitophilus zeamais. In an experiment to assess the insecticide action of organic extracts of Annona mucosa (Annonaceae), Petri dishes containing 10g of corn were treated with extracts prepared with different parts of the plant (seeds, leaves and branches) at a concentration of 1500mg/kg or just water (control), using a completely randomized design with 10 replicates. Then 20 Sitophilus zeamais adults were placed in each Petri dish and, after 60 days, the numbers of damaged and undamaged corn grains were counted, see Ribeiro et al (2013).

References

Moral, R. A., Hinde, J. and Demétrio, C. G. B. (2017) Half-normal plots and overdispersed models in R: the hnp package. Journal of Statistical Software 81(10):1-23.

Ribeiro, L. P., Vendramin, J. D., Bicalho, K. U., Andrade, M. S., Fernandes, J. B., Moral, R. A., Demétrio, C. G. B. (2013) Annona mucosa Jacq. (Annonaceae): A promising source of bioactive compounds against Sitophilus zeamais Mots. (Coleoptera: Curculionidae). J Stored Prod Res 55:6-14

Examples

data(corn)

# Binomial fit
model1 <- glm(cbind(y, m-y) ~ extract, family=binomial,
              data=corn)
anova(model1, test="Chisq")
hnp(model1, pch=4, main="Binomial: Logit",
    xlab="Half-normal scores", ylab="Deviance residuals")

# Quasi-binomial fit
model2 <- glm(cbind(y, m-y) ~ extract, family=quasibinomial,
              data=corn)
anova(model2, test="F")
summary(model2)$dispersion # estimated phi

# half-normal plots
par(mfrow=c(1,2),cex=1.4, cex.main=0.9, pty='s')
hnp(model1, pch=4, main="(a) Binomial; Logit",
    xlab="Half-normal scores", ylab="Deviance residuals")
hnp(model2, pch=4, main="(b) Quasibinomial; Logit",
    xlab="Half-normal scores", ylab="Deviance residuals")

anova(model1, test="Chisq") # binomial model
anova(model2, test="F") # quasi-binomial model
summary(model1) # binomial model
summary(model2) # quasi-binomial model

# now with factor level parameterisation
summary(update(model1,.~.-1))
summary(update(model2,.~.-1))

## for discussion on the analysis of this data set,
## see Demetrio et al. (2014)