[R-sig-ME] Proportion data: calculation of significance difference in infection rate

[Sophie Waegebaert]

Dear all,

I have a dataset with a MLPA analysis of 4-daily (0, 4, 8, 12, 16) sets of
samples of 20 individuals per treatment (control and DWV). I want to test
how the proportion of infected bees changed over time across treatment
conditions.

So, I have day 0 until day 16 and the amount of infected individuals per
treatment for each time point. After testing AIC and overdispersion, I made
a quasibinomial model (overdispersion parameter = 1.91) with day as a
continuous covariate:
fit_MLPA_quasi = glm(cbind(infected, not_infected)~treatment*day, family =
quasibinomial(), data = data)

I was wondering how I can determine until when the difference in infection
rate remained significant between the two treatments based on the 95%
confidence limits? I used the following command:

df = as.data.frame(effect(fit_MLPA_quasi, term="treatment:day"))
df

#   treatment day       fit        se      lower     upper
# 1      CTRL   0 0.1441225 0.6446699 0.03360447 0.4491729
# 2       DWV   0 0.9847726 1.6624012 0.52536941 0.9997354
# 3      CTRL   5 0.4521756 0.3817471 0.24490268 0.6774805
# 4       DWV   5 0.9745927 1.0791910 0.73229485 0.9981444
# 5      CTRL  10 0.8018179 0.4851205 0.55246817 0.9298731
# 6       DWV  10 0.9578982 0.7205049 0.79602594 0.9925174
# 7      CTRL  15 0.9519959 0.8273043 0.72370885 0.9933839
# 8       DWV  15 0.9310107 0.9093951 0.59317575 0.9920573

Besides not knowing how to calculate the significance, I do not understand
why R uses day 0, 5, 10 and 15, while in the dataset days 0, 4, 8, 12 and
16 are used.

Anyone who can give me a hint?

Thank you in advance.

Kind regards,
Sophie

	[[alternative HTML version deleted]]