dewlap_dfa: Discriminant Function Analysis (DFA) of dewlap color
In rscherrer/sagreicolor:

Description Usage Arguments Value Note Author(s)

This function performs multiple DFA on each island separately. The algorithm fits functions that best discriminate among habitats based on dewlap color data. The type of DFA can be linear (LDA) or quadratic (QDA). Each data point is then classified i.e. its original habitat is predicted based on the discriminant functions. Predictions can be jacknifed (leave-one-out). The significance of the classification is assessed in two ways. First, a MANOVA tests for differences in dependent variables across predicted groups. Second, a binomial test assesses the departure of the observed number of successful predictions from the null expectation.

1 2	dewlap_dfa(specdata, vars, type = "linear", plotit = T, CV = F, method = "bonferroni")

`specdata`	A data frame containing at least columns for the dependent variables, as well as a column "island" and a column "habitat".
`vars`	A character or integer vector. The names, or indices, of the dependent variables in `specdata`.
`type`	A character. Type of discriminant analysis. `"linear"` (default) or `"quadratic"`.
`plotit`	Logical. Whether to plot the loadings of the data points on the discriminant functions (applicable only if `type == "linear"`). Also, whether to plot adjusted P-values, across islands, of the MANOVAs and binomial tests.
`CV`	Logical. Whether to jacknife the predictions (cross-validation).
`method`	A character. The method to use for P-value correction.

A data frame with columns:

"observed", "expected", "n", "p.binom", "df", "wilks", "approx.F", "num.df", "denom.df", "p.manova"
observed: the observed number of succesful assignments
expected: the number of succesful assignments expected by change
n: the number of assignments
p.binom: the adjusted P-value of the binomial test
df: the degrees of freedom of the MANOVA
wilks: Wilk's lambda
approx.F: the approximate F-value calculated from Wilk's lambda
num.df: numerator degrees of freedom for F-test
denom.df: denumerator degrees of freedom for F-test
p.manova: adjusted P-value of the MANOVA

Quadratic discriminant analysis doesn't require homogeneous covariance matrices among groups, unlike linear (Robert I. Kabacoff, Quick-R, https://www.statmethods.net/advstats/discriminant.html).

Raphael Scherrer

rscherrer/sagreicolor documentation built on May 26, 2019, 12:32 p.m.