propr (version 3.1.8)

abstract: Abstract Two propr Objects

Description

This function abstracts a new propr object from two existing propr objects. The two propr objects should not have any overlapping samples. Typically, the two objects represent different experimental groups. The resultant abstracted object inherits all plot functions available for the original propr objects.

Usage

abstract(x, y, dt, colBy = "Adjusted", cutoff = 0.01)

Arguments

x, y

A propr object.

dt

A data.table. The result from prop2prob(x, y).

colBy

A character string. The column in dt used to select statistically significant pairs.

cutoff

A numeric scalar. The value of colBy used to select statistically significant pairs.

Value

Returns an abstracted propr object.

Details

The abstracted propr object has the following properties: The @counts and @logratio slots contain a join of the original slots via rbind. Meanwhile, the @matrix slot contains a difference matrix defined as tanh(atanh(x@matrix) - atanh(y@matrix)). The @pairs slot contains an index of all statistically significant pairs, toggled via the argument dt.

Visualizing the difference matrix with dendrogram may help summarize the results of prop2prob. Note that the difference matrix now also informs co-cluster assignment for the bucket, prism, and bokeh plots. Otherwise, most abstracted plots should match those made using perb(rbind(x@counts, y@counts)).

See Also

propr, prop2prob

Examples

# NOT RUN {
library(propr)
data(mail)
mail1 <- mail[1:2, ]
mail2 <- mail[3:4, ]
rho1 <- perb(mail1)
rho2 <- perb(mail2)
abstract(rho1, rho2)
# }