ModelChange {gRcox} | R Documentation |
Methods for altering RCOX models through tests. That is add the most significant or drop the most insignificant edge colour class, join the two most similar colour classes or split the most heterogeneous (composite) colour into atomic colour classes.
joinVCCTest(m, alpha = 0.05, statistic = c("wald", "dev"), criterion = c("test", "aic", "bic"), details = TRUE) joinECCTest(m, alpha = 0.05, statistic = c("wald", "dev"), criterion = c("test", "aic", "bic"), details = TRUE) joinCCTest(m, type = "ecc", alpha = 0.05, statistic = c("wald", "dev"), criterion = c("test", "aic", "bic"), details = TRUE) splitVCCTest(m, alpha = 0.05, criterion = c("test", "aic", "bic"), details = TRUE) splitECCTest(m, alpha = 0.05, criterion = c("test", "aic", "bic"), details = TRUE) splitCCTest(m, type = "ecc", alpha = 0.05, criterion = c("test", "aic", "bic"), details = TRUE) addECCTest(m, alpha = 0.05, criterion = c("test", "aic", "bic"), details = TRUE) dropECCTest(m, alpha = 0.05, statistic = c("wald", "dev"), criterion = c("test", "aic", "bic"), details = TRUE)
m |
An RCOX model object |
alpha |
Significance level |
type |
Either 'ecc' (edge colour class) or 'vcc' (vertex colour class) |
details |
If TRUE the details from the parameter comparisons are printed. |
statistic |
Use Wald or likelihood ratio test |
criterion |
Either 'test', 'aic' or 'bic' |
splitVCCTest and splitECCTest are simply wrappers for splitCCTest and likewise for joinVCCTest, joinECCTest and joinCCTest.
An RCOX model object.
Søren Højsgaard, sorenh@agrsci.dk
data(math) bufly0 <- rcox(vcc=list(~me, ~ve, ~al, ~an, ~st), ecc=list(~me:ve, ~me:al, ~ve:al, ~al:an, ~al:st, ~an:st), data=math) bufly1 <- rcox(~al:an:st, vcc=list(~me+st, ~ve+an), ecc=list(~me:ve+me:al, ~ve:al+al:st), data=math) joinECCTest(bufly0) splitECCTest(bufly1, criterion='aic') dropECCTest(bufly1,alpha=0.001,statistic='dev') addECCTest(bufly1)