Learn R Programming

nmm (version 0.9)

Nonlinear Multivariate Models

Description

Estimates a subset of nonlinear multivariate models (NMM): system of nonlinear regressions (SNR), logit, and a joint model of SNR and logit. 'nmm' uniquely accounts for correlations between the error terms from nonlinear regressions and the probabilities from logit models. It also enables a very flexible design of logit: alternative-specific indirect utilities, individual-specific choice set and number of actual choices.

Copy Link

Version

Install

install.packages('nmm')

Monthly Downloads

37

Version

0.9

License

GPL (>= 2)

Maintainer

Simona Jokubauskaite

Last Published

January 7th, 2021

Functions in nmm (0.9)

Add interactions

Returns function for sigma and rho calculation

Adjusted Akaike's Information Criterion.

add_variable adds columns to the data matrix

MNlogitf or MNdogitf returns log-likelihood(LL) expression for discrete equations of "logit" or "dogit" model.

LL_joint_no_corr

LL_joint_no_corr Function for log-likelihood without correlation between continuous and discrete equations

LL_joint Function for joint log-likelihood with correlation between continuous and discrete equations

Bread for Sandwiches.

Collapses lines ending with " "

check_names_start

Checks if starting values have names

check_par_av Checks if all elements of par are available in object.

combine_attr_deriv

combine_attr_deriv combines attributes of two deriv() results.

Example dataset

get_start_disc get starting values for discrete choice model.

ff_generate4maxle_p_v2

ff_generate4mlsem

Log-likelihood expressions for cont. equations

MAEDtimeExpenditure

Time-use and expenditure dataset

grad_hess_eval forms function of gradient and Hessian of log-likelihood produced by f_create.

formula2string removes square brackets from the supplied expressions. Convert par[2] -> par2, sigma[2] -> sigma_2_, sigma[2,3] -> sigma_2x2

Goodness of fit measures

convert_attr2exp

convert_attr2exp converts symbolic attribute of derivative into expression object.

generate_disc_coefs

Function for finding starting values for nls equations with errors

Significance of correlation matrix

get_npar Get number of parameters or vector of parameters in supplied equations. Extracts the number of parameters used in equations. Parameters are given as par[1], ..., par[n].

Goodness of fit measures for both parts

deriv_disc_expr_corr

Hessian and Gradient expressions

dat4cond_mean_cov_expr

Log-likelihood expressions for cont. equations plus 1 discrete

in2nmm convert some estimation results into nmm object.

get_par replaces names of parameters with par[i].

Gradient with supplied coefficients

f_create creates functions for log-likelihood of different models.

deriv_disc_expr

Get expression of Hessian and Gradient for discrete choice model with quantile transformation

Log-likelihood expressions for cont. equations sem

get_start_cont get starting values for continuous equations.

Internal helper function

get_start get starting values for discrete or continuous choice model.

mlsem returns expression of log-likelihood for joint normal distribution, for maximum likelihood (ML), Simultaneous Equations Models (SEM) variant.

Meat for Sandwiches

Maximum likelihood estimation of nonlinear multivariate models (NMM).

Modifies function to optimize sigma and rho

nrParam return number of parameters used in the estimation.

Converts vector of variances and correlations values into a matrix

Log-likelihood expressions for cont. equations

expr_ll_norm_v2

Another Log-likelihood expressions for cont. equations version 2

ff_generate4maxle_p

Internal helper function for MNdogitf

searches for 1st good optimization method between possible maxLik algorithms

Hessian with supplied coefficients

ff_generate4maxle

searches for 1st good optimization method between possible maxLik algorithms and then applies DEoptim

replace_expression

Replaces expressions ".exprXX" with strings

replace_par replaces text with other text.

Calculate Heat rate

wxMaxima does symbolic computation in 'Maxima' Requires installation of Maxima software.

help_diagnostics

Generates objects needed for diagnostics functions

cond_expr returns moments of conditional multivariate normal distribution X|Y (last variable is dependent). Only expression for X|Y. Requires installation of Maxima software.

extract_attr_deriv

extract_attr_deriv converts attributes(hessian/gradient) of deriv() into a matrix of character strings.

maxle returns expression of log-likelihood (LL) of joint normal distribution.

cond_mean_cov_expr

cond_mean_cov_expr returns conditional mean and variance X|Y of conditional multivariate normal distribution . X is discrete variable, Y are continuous variables. No correlation between discrete eq.

extract_cont_data

Takes out only data used in continuous estimation

replace_par_wrap

replace_par_wrap replace text with other text, wrapper of replace_par.

Log-likelihood(LL) with supplied coefficients.

Converts sigma matrix into parameter vector

Internal helper function for MNlogitf

maxle_p returns expression of partitioned log-likelihood. f(y1,y2,..,yn)=f(y1)f(y2|y1)f(y3|y2y1)...f(yn|y1..y(n-1))

prepare_data prepare data for the estimation.

nmm_lastDEoptim

last time try out DEoptim if no functions before produced "good" results

Helper functions

searches for best optimization method between possible maxLik algorithms

Test if maxLik produces an error

Checks if nmm object has an error

string2formula add square brackets to expressions. Reverse of formula2string. Convert par[2] <- par2, sigma[2] <- sigma_2_, sigma[2,3] <- sigma_2x2