Standard univariate statistics for columns of a Filebacked Big Matrix. For now, the sum and var are implemented (the mean and sd can easily be deduced, see examples).

big_colstats(X, ind.row = rows_along(X), ind.col = cols_along(X))

Arguments

X

An object of class FBM.

ind.row

An optional vector of the row indices that are used. If not specified, all rows are used. Don't use negative indices.

ind.col

An optional vector of the column indices that are used. If not specified, all columns are used. Don't use negative indices.

Value

Data.frame of two numeric vectors sum and var with the corresponding column statistics.

See also

Examples

set.seed(1) X <- big_attachExtdata() # Check the results str(test <- big_colstats(X))
#> 'data.frame': 4542 obs. of 2 variables: #> $ sum: num 680 821 789 843 562 666 902 537 536 553 ... #> $ var: num 0.46 0.324 0.393 0.311 0.518 ...
# Only with the first 100 rows ind <- 1:100 str(test2 <- big_colstats(X, ind.row = ind))
#> 'data.frame': 4542 obs. of 2 variables: #> $ sum: num 138 157 115 181 92 106 167 124 120 94 ... #> $ var: num 0.46 0.369 0.513 0.176 0.579 ...
plot(test$sum, test2$sum)
abline(lm(test2$sum ~ test$sum), col = "red", lwd = 2)
X.ind <- X[ind, ] all.equal(test2$sum, colSums(X.ind))
#> [1] TRUE
all.equal(test2$var, apply(X.ind, 2, var))
#> [1] TRUE
# deduce mean and sd # note that the are also implemented in big_scale() means <- test2$sum / length(ind) # if using all rows, # divide by nrow(X) instead all.equal(means, colMeans(X.ind))
#> [1] TRUE
sds <- sqrt(test2$var) all.equal(sds, apply(X.ind, 2, sd))
#> [1] TRUE