Description Usage Arguments Details Value Note Author(s) References See Also Examples
Calculates repeatability from a generalised linear mixed-effects models fitted by PQL (penalized-quasi likelihood) estimation for binary and proportion data.
1 | rpt.binomGLMM.multi(y, groups, link=c("logit", "probit"), CI=0.95, nboot=1000, npermut=1000)
|
y |
Vector of a response values (for binary data) or a two-column matrix, array or data.frame with colums m, n-m, where m is the number of successes and n the number of trials. |
groups |
Vector of group identitites. |
link |
Link function, |
CI |
Width of the confidence interval (defaults to 0.95). |
nboot |
Number of parametric bootstraps for interval estimation (defaults to 1000). Larger numbers of permutations give a better asymtotic CI, but may be very time-consuming. |
npermut |
Number of permutations for significance testing (defaults to 1000). Larger numbers of permutations give better asymtotic P values, but may be very time-consuming. |
Models are fitted using the glmmPQL function in MASS with the quasibinomial
family (proportion data) or the binomial
family (binary data).
Returns an object of class rpt that is a a list with the following elements:
datatype |
Type of response (here: "binomial"). |
method |
Method used to calculate repeatability (here: "PQL"). |
link |
Link function used (here: "logit" or "probit"). |
CI |
Width of the confidence interval. |
R.link |
Point estimate for repeatability on the link scale. |
se.link |
Standard error (se) for repeatability on the link scale, i.e. the standard deviation of the parametric bootstrap runs. Note that the distribution might not be symmetrical, in which case the se is less informative. |
CI.link |
Confidence interval for repeatability on the link scale based on parametric-boostrapping of R. |
P.link |
Approximate P value from a significance test for the link scale repeatability based on randomisation. |
R.org |
Point estimate for repeatability R on the original scale. |
se.org |
Standard error (se) for repeatability on the original scale, i.e. the standard deviation of the parametric bootstrap runs. Note that the distribution might not be symmetrical, in which case se is less informative. |
CI.org |
Confidence interval for repeatability on the link scale based on parametric-boostrapping of R. |
P.org |
Approximate P value from a a significance test for the original scale repeatability based on randomisation. |
omega |
Multiplicative overdispersion parameter. |
R.boot |
Named list of parametric bootstap samples for R. |
R.permut |
Named list of permutation samples for R. |
Confidence intervals and standard errors are inappropriate at high repeatabilities (omega < 1), because parametric bootstrapping allows only omega greater than or equal to 1.
Holger Schielzeth (holger.schielzeth@ebc.uu.se) & Shinichi Nakagawa (shinichi.nakagawa@otago.ac.nz)
Browne, W. J., Subramanian, S. V., et al. (2005). Variance partitioning in multilevel logistic models that exhibit overdispersion. Journal of the Royal Statistical Society A 168: 599-613.
Goldstein, H., Browne, W., et al. (2002). Partitioning variation in multilevel models Understanding Statistics 1: 223-231.
Nakagawa, S. and Schielzeth, H. (2011) Repeatability for Gaussian and non-Gaussian data: a practical guide for biologists. Biological Reviews 85: 935-956.
rpt.binomGLMM.add, rpt, print.rpt
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 | ## Not run:
# repeatability estimations for egg dumping (binary data)
data(BroodParasitism)
attach(BroodParasitism)
(rpt.Host <- rpt.binomGLMM.multi(HostYN[OwnClutchesBothSeasons==1], FemaleID[OwnClutchesBothSeasons==1],
nboot=10, npermut=10)) # low number of nboot and npermut to speed up error checking
(rpt.BroodPar <- rpt.binomGLMM.multi(cbpYN, FemaleID, nboot=10, npermut=10))
# low number of nboot and npermut to speed up error checking
detach(BroodParasitism)
# repeatability estimations for egg dumping (proportion data)
data(BroodParasitism)
attach(BroodParasitism)
ParasitisedOR <- cbind(HostClutches, OwnClutches-HostClutches)
(rpt.Host <- rpt.binomGLMM.multi(ParasitisedOR[OwnClutchesBothSeasons==1,],
FemaleID[OwnClutchesBothSeasons==1], nboot=10, npermut=10)) # reduced number of npermut iterations
ParasitismOR <- cbind(cbpEggs, nEggs-cbpEggs)
zz = which(ParasitismOR[,1]==0 & ParasitismOR[,2]==0) # some rows have entries 0,0 and need to be removed
(rpt.BroodPar <- rpt.binomGLMM.multi(ParasitismOR[-zz,], FemaleID[-zz], nboot=10, npermut=10))
# reduced number of npermut iterations
detach(BroodParasitism)
## End(Not run)
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.