Options:

date

produces string containing the date and time execution completed and the ${\mathfrak{A}melia}$ version number

means

creates (iteration x _AMp) matrix of means where iteration is the length of the burn-in for IP, SIP, or VIP and _AMp is the number of variables in the dataset

cov

creates (iteration*_AMp x _AMp) matrix of variances/covariances where iteration is the length of the burn-in for IP, SIP,or VIP and _AMp is the number of variables in the dataset.

covimp

creates (_AMnds * _AMgap * _AMp x _AMp) matrix of variances and covariances where _AMnds is the number of datasets created by the imputation procedures, _AMgap is the number of iterations for IP, SIP, and VIP procedures between any two concurrent final imputations, and _AMp is the number of variables in the dataset. This is the time series matrix of variances and covariances from the imputation stage.

llike

creates a vector of log-likelihoods which is equal to the length of the burn in for IP, SIP, and VIP and equal to the number of iterations needed for convergence for EM

llikeimp

creates vector of log-likelihoods which is equal to the length of _AMnds * _AMgap, the total number of iterations in the impute stage for IP, SIP, and VIP. This is the log-likelihood vector for the imputation stage.

lpost

creates a vector of log-posterior densities which is equal to the length of the burn in for IP, SIP, and VIP and equal to the number of iterations to convergence for EM

meansimp

creates (_AMnds * _AMgap x _AMp) matrix of means where _AMnds is the number of datasets created by the imputation procedures, _AMgap is the number of iterations for IP, SIP, and VIP procedures between any two concurrent final imputations, and _AMp is the number of variables in the dataset. This is the time series matrix of means from the imputation stage.

resamp

produces the number of resamples in the importance sampling for EMis

thetafin

produces the final stacked, centered, and scaled theta matrix

dffin

produces the final degrees of freedom parameter associated with thetafin when the $t$-distributed model is implemented _AMemt=1.

wfin

produces the final vector of weights associated with thetafin and dffin when the $t$-distributed model is implemented _AMemt=1.

thetasav

produces the matrix with saved thetas, if they were saved

thetan

the value of theta used in the n-th imputation.

dfn

the value of the degrees of freedom used in the n-th imputation (for _AMemt=1).

weightn

the value of weights used in the n-th imputation (for _AMemt=1).

Amelia globals

The values of all external globals from the amelia procedure can also be read. (The following globals are internal and output globals from amelia.)

_AMdat

scaled and stacked dataset used by algorithms

_AMf

= (_AMn x _AMp) indicator matrix of dataset: 1 if in the stack; 0 if not in the stack

_AMi

(s+1 x 1) matrix that indicates which line of the stacked data, the missingness pattern s is first found. That is, i[n] would give the row number for the first row with the nth missingness pattern. Also, since there is no s+1 missingess pattern, the last element is defined as: i[s+1]=n+1

_AMimpmk

0 if marker indicating that buffer has not recorded an imputation, 1 if such a run has been recorded in the buffer

_AMlongo

(_AMp x s) matrix recording the columns of observed data for each pattern of missingness

_AMmu

means of observed data

_AMn

number of observations in dataset

_AMnod

records initial row number of each row in the stacked data

_AMorigp

equals number of variables in original dataset, is equal to _AMp, the number of variables for the imputation procedure, if there are no nominal variables in the dataset.

_AMorigv

matrix of the observed nominal variables

_AMp

number of variables in dataset.

_AMpod

records initial column number of each column in the stacked data.

_AMr1

(_AMn x _AMp) stacked and sorted indicator matrix for dataset; 1 if observed, 0 if not observed.

_AMr1o

(_AMn x _AMp) indicator matrix for original dataset; 1 if observed, 0 if not observed.

_AMr2

(s x _AMp) indicator matrix 1 if observed 0 if not observed where s is the number of unique missingness patterns in the dataset.

_AMstd

standard deviations of observed data.

_AM1stln

1 if marker indicating 1st missingness pattern completely observed; 0 if marker indicating no missingness pattern completely observed.

_AMx0s

(_AMn x _AMp) stacked dataset with each missing value code ``.'' replaced by a numeric zero.

_AMvers

string designating version of ${\mathfrak{A}melia}$ program.

_AMcopy

copy of original dataset if time-series-cross-section features are used.

_AMcso

copy of original vector identifying multiple cross-sections for time-series-cross-section features.

_AMnewtv

identifies actual position of lagged variables as entered into the imputation model.