setinitvalues: The function to set initial values for parameters:...

Description Usage Arguments Value Examples

View source: R/setinitialvalues.R

Description

Generate initial values for parameters

Usage

1
2
3
4
setinitvalues(npred, np, npred_miss, npred_sub, nmiss, nsid,
    censor_lim_upp = 0.008, ita_a = 1, ita_b = 1/10, g_mu = 0,
    g_sig = 1, alpha_mu_u = 0, alpha_mu_s = 1, alpha_theta_a = 1,
    alpha_theta_b = 1/10, beta2_theta_a = 1, beta2_theta_b = 1/10)

Arguments

npred

number of predictors for the regression model

np

number of protein/metabolite units comprised of the response values (i.e. which represents peptides' ion-intensities used to construct protein/metabolite's abundance)

npred_miss

number of predictors for missingness

npred_sub

number of predictors for the second level units such as subjects

nmiss

number of observations with missing responses values

nsid

number of second level units i.e. subjects

censor_lim_upp

upper-limit of censored value of responses. The default value is 0.001 according to an experiment device. User can change it according to the data.

ita_a

shape parameter for gamma distributed prior-ita (std of response value). The default is set to 1.

ita_b

rate parameter for gamma distributed prior-ita. The default is set to 1/10. The default values of shape and rate parameters provide a reasonable wide range of initial value for ita. Users can change it accordingly.

g_mu

mean of normal distributed location parameter g for re-parameterising U (regression coefficient of model in the completed data). The default value is set to 0 for the mean of standard normal distribution.

g_sig

std of normal distributed location parameter g. The default is set to 1 for the std of normal distribution.

alpha_mu_u

mean of normal distributed location parameter alpha_mu for re-parameterising alpha (regression coefficient of logistic regression model for missing prob). The default is set to 0.

alpha_mu_s

std of normal distributed location parameter alpha_mu

alpha_theta_a

shape parameter of gamma-distributed dispersion parameter alpha_theta for re-parameterising alpha. Default value is set to 1 as a natural starting value.

alpha_theta_b

rate parameter of gamma-distributed alpha_theta. Default value uses 1/10 as for ita_b. Both default values of shape(_a) and rate(_b) of alpha_theta can be changed to give a wider range (_b=1/10) or a narrower range (_b=0.5).

beta2_theta_a

shape parameter of gamma-distributed beta2_theta for re-parameterizing beta2 (regression coefficient for second level units,i.e. subject). Default value uses 1.

beta2_theta_b

rate parameter of gamma distributed dispersion parameter beta2_theta. Default value used 1/10, same as for ita_b.

Value

pVAR precision matrix for predictors in completed data model

U_latent standardized multinormal distributed latern variable to re-parameterise regression coefficient U.

g location parameter to re-parameterise U.

alpha_mu mean value for alpha(regression coefficient of model for missing probability).

alpha_latent standardized normal distributed latent variable to re-parameterize alpha.

beta2_latent standardized multivariate normal distributed latent variable to re-parameterising beta2.

beta2_mu mean of the multivariate normal distributed beta2

y_m_latent standardized normal distributed latent variable to re-parameterise response variable.

Examples

1
2
testexmp <- setinitvalues(npred=2,np=3,npred_miss=3,npred_sub=2,nmiss=10,
nsid=30)

mlm4omics documentation built on Oct. 31, 2019, 9:43 a.m.