LmImputeMulti: INTERNAL FUNCTION: Repeated regression imputation.

In statisticsnorway/Kostra: Functions for Kostra

INTERNAL FUNCTION: Repeated regression imputation.

Description

LmImpute is run several times using different versions of input parameters. Output from each run of LmImpute is in output of this function. In each run it is possible to update the interest variable (y) with available imputed values. Thus the function can be used to impute using a primary x-variable and a secondary x-variable for cases where the primary is missing.

Usage

LmImputeMulti(
  data,
  yName = "y",
  xModel = c("x1", "x2"),
  yModel = yName,
  weights = NULL,
  limitModel = 2.5,
  limitIterate = 4.5,
  limitImpute = 50,
  maxiter = 10,
  returnIter = TRUE,
  returnYHat = FALSE,
  returnFirst = FALSE,
  returnLast = TRUE,
  returnFinal = FALSE,
  MultiFuction = function(x) {
     max(abs(x))
 },
  estimationGroup = TRUE,
  unfoldCoef = FALSE,
  BackTransform = list(NULL),
  warningEstimate = vector("list", n),
  removeEmpty = FALSE,
  replaceByImputed = TRUE,
  imputedInModel = FALSE,
  category123FromFirst = FALSE,
  cvPercent = TRUE,
  returnSameType = FALSE
)

LmImpute2(
  data,
  warningEstimate = list(NULL, "estimate: Missing yImputed replaced by zero"),
  replaceByImputed = TRUE,
  cvPercent = TRUE,
  ...
)

LmImputeOne2Many(
  data,
  yName = "y",
  xModel = "x",
  yModel = c(yName, "yMany"),
  category123FromFirst = TRUE,
  keepSinge = NULL,
  keepMulti = NULL,
  ...
)

Arguments

data	Input data set (data.frame, data.table or list)
yName	Name of interest variable in data set
xModel	Vector of strings with the right part model formula.
yModel	String with left part model formula (vector possible)
weights	NULL or string with weight expression (vector possible)
limitModel	Studentized residuals limit. Above limit -> category 2.(vector possible)
limitIterate	Studentized residuals limit for iterative calculation of studentized residuals.(vector possible)
limitImpute	Studentized residuals limit. Above limit -> category 3. No imputation when 0.(vector possible)
maxiter	Maximum number of iterations.(vector possible)
returnIter	When TRUE, iteration when observation was thrown outin output.(vector possible)
returnYHat	When TRUE, fitted values and corresponding estimates in output. (vector possible)
returnFirst	When TRUE, studentized residuals from first iteration in output. (vector possible)
returnLast	When TRUE, some results from last iteration in output.(vector possible)
returnFinal	When TRUE, extra results from final model in output. (vector possible)
MultiFuction	Transforming rStud for several responses into a single positive value.
estimationGroup	Total estimates will be computed within each group. Default (and TRUE) is a single group (estimationGroup <- rep(1, N) ).
unfoldCoef	When TRUE several elements of coef will be spilt as several ouput elements.
BackTransform	When model contains transformation of y (e.g: "log(y)~x") a function (e.g: exp) can be supplied to transform back to original scale before calculation of leaveOutResid, yHat, yImputed, estimate, estimateYHat, estimateOrig and seRobust. (list possible)
warningEstimate	Warning text when missing values. Use NULL to avoid warning. (list possible)
removeEmpty	When TRUE empty elements will be removed from output.
replaceByImputed	When TRUE missing values of the interest variable (y) is replaced by imputed values in each round.
imputedInModel	When FALSE above imputed values are omitted from subsequent models (category 2 forced).
category123FromFirst	Whene TRUE category123 from first run is input to subsequent LmImpute calls.
cvPercent	When TRUE (default) cv output is in percent
returnSameType	When TRUE and when the type of input y variable(s) is integer, the output type of yImputed and estimate is also integer. Estimates/sums are then calculated from rounded imputed values.
keepSinge	When non-NULL only output elements with names in keepSinge are kept from first LmImputeOne2Many run.
keepMulti	When non-NULL only output elements with names in keepMulti are kept from second LmImputeOne2Many run.

Details

LmImputeMulti performs several calls to LmImpute and the number of calls is the length of xModel. Other parameters can also change between calls by specifying them as vectors or lists.

LmImpute2 is a specialized variant for two LmImpute runs only and combined estimates of seRobust, seEStimate and cv are calculated when replaceByImputed = TRUE.

LmImputeOne2Many is another specialized variant meant for two runs using the same model except that the first run has a single y and the next run several y's. Category123 from the first run is used in the second.

Value

Output of LmImputeMulti is a list where each element is the output of LmImpute. Output of LmImpute2 is not such a list. Instead the names are changed using "A" and "B". Output of LmImputeOne2Many is not such a list. Instead the names from second round are changed using "M".

Examples

# -----  LmImpute2 and LmImputeMulti -----
set.seed(123)  # same results each time
z = data.frame(  # Small test data set
x1 = c(NA,2:19,NA),
x2 = rep(1:20),
y = runif(20)+c(rep(0,15),NA,1E3,1E5,NA,NA))
LmImpute2(z)
LmImputeMulti(z)
LmImputeMulti(z,xModel=c("x1","x2","x2"),limitImpute=c(50,50,3), replaceByImputed=FALSE)


 # -----  LmImputeOne2Many -----
 # Create data fram z with matrix embedded in one variable
 # The single y is also the first variable in the matrix yMany
 set.seed(123)  # same results each time
 y = runif(20)+c(rep(0,15),NA,1E3,1E5,NA,NA)
 z = data.frame(  # Small test data set
   x = c(1,1,1:10,3,3,3,4,4,5,5,5),
   y = y,
   yMany = I(cbind(y,matrix(1:60,20,3,dimnames=list(NULL,c("A","B","C"))))))
 a=LmImputeOne2Many(z,limitModel=1)
 print(a)
 a$MrStud[, 1] -  a$rStud # Not equal since no iteration and no "last" when multiple y
 a=LmImputeOne2Many(z,limitModel=1,returnFinal=TRUE)
 a$MrStud[, 1] -  a$rStud # Now equal

statisticsnorway/Kostra documentation built on Sept. 25, 2024, 10:37 a.m.