imi.t.test.more: imi.t.test.more
In vahidnassiri/imi: Implemenets Iterative Multiple Imputation

Description Usage Arguments Value References Examples

This function uses the function mi.t.test in R package MKmisc to perform one or two sample Student's t-test on an incomplete dataset using multiple imputation and determines the sufficient number of imputations using iterative multiple imputation (imi) procedure.

1
2
3

imi.t.test(data.miss,data.imp0,max.M=500,epsilon,method='pmm',
                       x, y = NULL, alternative='two.sided',mu,paired = FALSE, var.equal = FALSE,
                       conf.level = 0.95,conv.plot=TRUE,successive.valid=3)

`data.miss`	A data frame with the variable in the model as its columns. Note that the missing values should be indicated by NA.
`data.imp0`	A list with already imputed sets of data as its components.
`max.M`	The maximum number of iterations which the algorithm should terminate afterwards in case of non-covergence.
`epsilon`	The threshold for difference between two iterations.
`method`	Specifying string value 'mvn' would impute the data using a multivariate normal predictive model in R package amelia2, any other specification will impute the data using fully conditional specification approach in R package mice. One can see the method in documentation of function mice in R package mice. Specifying 'auto' will selected the predictive model based on the measurement level of each variable.
`x`	Name of the data column which the t-test should be performed on it.
`y`	In case of two sample t-test this input variable specifiess the name of the second column for paired-test, or an indicator variable showing different populations for indepdendent two sample t-test.
`alternative`	A character string specifying the alternative hypothesis, it takes the values "two.sided" (default), "greater" or "less".
`mu`	A real number specifying the test value.
`paired`	A logical value specifying whether a paired t-test should be performed or not.
`var.equal`	A logical value specifying whether variance equality should be assumed or not.
`conf.level`	Confidence level of the test.
`conv.plot`	A logitical value, if TRUE then a convergence plot will be generated, if FALSE no plot will be provided.
`successive.valid`	An integer with minimum 1 which specifies the number of successive steps the stopping rule should be validated so the procedure could terminate.
`print.progress`	A logical variable, if TRUE it prints the progress of imputation.

`test.result`	A list with the final MI-based outcome of the t-test.
`data.imp`	A list with imputed datasets as its components.
`dis.steps`	A vector with computed distance between iterations.
`conv.status`	If 1 then convergence is achieved, if 0 with max.M iterations, still the convergence could not be achieved.
`M`	The selected number of imputed datasets.

https://stat.ethz.ch/R-manual/R-devel/library/stats/html/t.test.html

# To illustrate the use of this function we use the cholestrol dataset in R package norm2.
> library(norm2)
> data(cholesterol)
>
> chol.ttest=imi.t.test (cholesterol,M0='manual',max.M=500,epsilon=0.05/10,method='mvn',
+                           x=names(cholesterol)[3], y = NULL, alternative='two.sided',mu=220,
+                           paired = FALSE, var.equal = FALSE,
+                           conf.level = 0.95,conv.plot=TRUE,successive.valid='manual')
-- Imputation 1 --

  1  2  3  4  5  6

[1] "The time it takes (in seconds) to imput the data once and fit the model to it is: 0.01"
What is your choice of initial number of imputations?2
What is your choice for successive steps validation?3
-- Imputation 1 --

  1  2  3  4  5  6  7  8

-- Imputation 1 --

  1  2  3  4  5  6  7

# it continues till the convergence happens.