Where the manual's remaining commands live in R

Working through TDA's manual section by section, some commands have no function of their own here. None of them is missing functionality; each is covered in one of three ways.

Details

Subsumed by R itself. The matrix definition commands exist because TDA has no other way to get a matrix into a session; R does. mdef (literal) is matrix(); mdefc (constant) is matrix(d, m, n); mdefi (rectangular identity) is diag(1, m, n); mdeff (from file) is as.matrix(read.table(...)) or tda_read_table; mpr (print a matrix expression) is printing the return value of any matrix wrapper. The expression operators of section 5.2.5 – nd, ndi, td, chd, fd, mr, ndf, poisson, negbin, bivn, jul, july, julm, juld and relatives – are pnorm, qnorm, pt, pchisq, pf, the Mill's ratio dnorm(x)/(1 - pnorm(x)), dnorm, dpois/dnbinom (TDA returns the log density), mvtnorm::pmvnorm, and R's date arithmetic. The engine's implementations of all of these are compared against R's on every run of the differential fuzzer, so using either side gives the same numbers; to evaluate one of TDA's operators directly, pass the expression to tda_evalf.

Delivered by another wrapper. mdefg (adjacency matrix from the current graph) is tda_g_edges(g, format = "full_square"), with -1 marking absent edges. mbrr is tda_brr. mple is tda_mple. mlp's raw-tableau form is assembled internally by tda_mlp from the objective, constraints, and bounds.

Reached through the session they belong to. mdefb (a block of the data matrix) requires a dblock context and is used through the commands= argument of tda_dblock. mplog, mppar, mpcov and mpgrad are parameters OF estimation commands, not commands; their contents are what the fitted objects already carry as $logLik, $estimates and $vcov (case-specific gradients, mpgrad, can be requested through a command's options= where supported).

Chapter 5 names that are not standalone commands:

mplog, mppar, mpcov, mpgrad

parameters of the estimation commands that write the log likelihood, parameters, covariance matrix or gradient into a named TDA matrix. The R wrappers return the same quantities directly ($logLik, $estimates, $vcov), so these are only needed inside a hand-written tda_run matrix pipeline, where they pass through options= as usual.

mdef, mpr, mdefc, mdefi, mdeff

matrix definition and printing, which R does natively: matrix(...), print(), matrix(d, m, n), diag(1, m, n) and as.matrix(read.table(fname)); every matrix wrapper accepts plain R matrices. mdefg (a graph's adjacency matrix) does compute something and has its wrapper, tda_mdefg.

mdefb

extracts one dblock-defined block of the data matrix into a matrix – in R, subset the data frame; in a tda_run script it works after a dblock command as documented.

mbrr

the manual's name for brr; see tda_brr.

sort (command)

see tda_esort and the sort= option of pdata in tda_read_sys.

lag, suc, snum, rank, sort (operator)

operators of TDA's expression language, used inside nvar variable definitions – e.g. L[6.0]=lag(Y,1), R[6.0]=rank(Y) – in a tda_run script. snum, rank and sort take the variable as an argument.