hnp (version 1.2-7)

chryso: Chrysoperla externa mortality data

Description

Mortality of the predator Chrysoperla externa on different doses of lime sulphur, a substance used to control pests on trees.

Usage

data(chryso)

Format

A data frame with 24 observations on the following 4 variables.

dead	numeric	count of dead C. externa specimens
alive	numeric	count of alive C. externa specimens
conc	numeric	lime sulphur concentration
log.conc	numeric	natural logarithm of lime sulphur concentration

Details

The neuropteran Chrysoperla externa is a predator that acts as a natural enemy of the brown citrus aphid, Toxoptera citricida, which is among the most important citrus pests worldwide. A possible strategy to control T. citricida populations would be to use a substance called lime sulphur and the C. externa predator in combination, which may be beneficial as long as the lime sulphur has less effect on the predator than the prey. To explore this, Battel (2012) conducted an experiment with first-instar larvae of Chrysoperla externa exposed to different levels of lime sulphur. Specifically, twenty-four Orange Jessamine (Murraya paniculata) plants were sprayed with different concentrations (conc) of lime sulphur and up to seven first-instar larvae were placed on each plant. The experiment was set up in a completely randomized design with four treatments: lime sulphur concentrations at 0ppm (water control), 60ppm, 600ppm, and 6000ppm. The plants were observed until the predators reached the second instar and the number of larvae that died on each plant was recorded.

References

Battel, A. P. M. B. (2012) Dinamica de predacao e resposta funcional em Chrysoperla externa (Neuroptera: Chrysopidae) sobre Toxoptera citricida (Hemiptera: Aphididae) aplicada a citricultura organica. Master's dissertation, ESALQ-USP

Examples

data(chryso)

# fit model using conc levels both on log-scale and
# as a factor to produce simple analysis of deviance
model <- glm(cbind(dead, alive) ~  log.conc + factor(conc),
             family=binomial, data=chryso)
anova(model, test="Chisq")
# test adequacy of factor model using deviance and X2
1-pchisq(deviance(model), df.residual(model))
(X2 <- sum(residuals(model, type="pearson")^2))
1-pchisq(X2, df.residual(model))

model1 <- glm(cbind(dead, alive) ~ log.conc,
              family=binomial, data=chryso)
par(mfrow=c(1,2), cex=1.4)
with(chryso, plot(jitter(log.conc), dead/(dead+alive),
                  ylab="Proportion dead", xlab="log(conc+1)"))
x <- seq(0, 8.7, .1)
pr <- predict(model1, data.frame(log.conc=x), ty="response")
lines(x, pr)
# half-normal plot
hnp(model1, xlab="Half-normal scores",
    ylab="Deviance residuals", pch=4)

require(MASS)
dose.p(model1, p=.10)

logLC10 <- dose.p(model1, p=.10)
LC10 <- exp(logLC10[[1]]) - 1

#95% CI on log-dose scale using transformation
c(logLC10[1]-2*attr(logLC10,'SE'),
  logLC10[1]+2*attr(logLC10,'SE'))

#95% CI on dose scale using transformation
c(exp(logLC10[1]-2*attr(logLC10,'SE'))-1,
  exp(logLC10[1]+2*attr(logLC10,'SE'))-1)

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