MaxContribution {GaussSuppression}

Find major contributions to aggregates

Description

Assuming aggregates are calculated via a dummy matrix by z = t(x) %*% y, the n largest contributions are found (value or index) for each aggregate.

Usage

MaxContribution(
  x,
  y,
  n = 1,
  decreasing = TRUE,
  index = FALSE,
  groups = NULL,
  return2 = FALSE
)

Arguments

x

A (sparse) dummy matrix

y

Vector of input values (contributors)

n

Number of contributors to be found

decreasing

Ordering parameter. Smallest contributors found when FALSE.

index

Indices to y returned when TRUE

groups

When non-NULL, major contributions after aggregation within groups. Cannot be combined with index = TRUE. The missing group category is excluded.

return2

When TRUE, two matrices are returned, value and id. The latter contains indices when group is NULL and otherwise a character matrix of groups.

Value

Matrix with lagest contributions in first column, second largest in second column and so on. Alternative output when using parameters index or return2.

Author(s)

Øyvind Langsrud

See Also

ModelMatrix

Examples

library(SSBtools)

z <- SSBtoolsData("sprt_emp_withEU")
z$age[z$age == "Y15-29"] <- "young"
z$age[z$age == "Y30-64"] <- "old"

a <- ModelMatrix(z, formula = ~age + geo, crossTable = TRUE)

cbind(as.data.frame(a$crossTable), MaxContribution(a$modelMatrix, z$ths_per, 1))
cbind(a$crossTable, MaxContribution(a$modelMatrix, z$ths_per, 10))
cbind(a$crossTable, MaxContribution(a$modelMatrix, z$ths_per, 10, index = TRUE))

# Both types of output can be achieved with return2 = TRUE)
identical(MaxContribution(a$modelMatrix, z$ths_per, 10, return2 = TRUE),
          list(value =  MaxContribution(a$modelMatrix, z$ths_per, 10), 
               id =  MaxContribution(a$modelMatrix, z$ths_per, 10, index = TRUE)))

b <- ModelMatrix(z[, -4], crossTable = TRUE, inputInOutput = c(TRUE, FALSE, TRUE))

k <- cbind(b$crossTable, MaxContribution(b$modelMatrix, z$ths_per, 10))

gr18 <- paste0("g", 1:18)                          # Each row is a group
k18 <- cbind(b$crossTable, MaxContribution(b$modelMatrix, z$ths_per, 10, groups = gr18))
identical(k, k18) # TRUE

gr9 <- paste0("g", as.integer(10 * z$ths_per)%%10) # 9 groups from decimal
k9 <- cbind(b$crossTable, MaxContribution(b$modelMatrix, z$ths_per, 10, groups = gr9))

k18[c(4, 13, 17, 33), ]
k9[c(4, 13, 17, 33), ]

# Group info obtained with return2 = TRUE
k9_id <- cbind(b$crossTable, MaxContribution(b$modelMatrix, z$ths_per, 10, groups = gr9, 
                                             return2 = TRUE)$id)
k9_id[c(4, 13, 17, 33), ]


# Verify similarity
z$y <- z$ths_per + (1:nrow(z))/100  # to avoid equal values
id1 <- MaxContribution(b$modelMatrix, z$y, 10, index = TRUE)
id1[!is.na(id1)] <- paste0("g", id1[!is.na(id1)])
mc2 <- MaxContribution(b$modelMatrix, z$y, 10, groups = gr18, return2 = TRUE)
id2 <- mc2$id
identical(id1, id2)