Reconstruct a tbl_sample after dplyr operations — dplyr_reconstruct.tbl

This method ensures that the tbl_sample class is only preserved when the essential column (.weight) is still present. Operations like summarise() or count() that remove this column will return a regular tibble instead.

Usage

# S3 method for class 'tbl_sample'
dplyr_reconstruct(data, template)

Arguments

data: The modified data
template: The original tbl_sample object

Value

A tbl_sample if essential columns present, otherwise a tibble

Details

Operations that change the number of rows (filtering, joins that drop or duplicate rows) keep the class but mark the sample as modified: the stored design then no longer describes the data, so design-based computations (as_svydesign(), joint_expectation(), design_effect()) refuse the sample and point to survey-side domain analysis instead. Dropping an internal design column (for example select() without .fpc_1) is marked the same way, because the export would silently change (a missing FPC column falls back to no finite population correction). Reordering rows, reordering columns, and adding ordinary data columns are harmless and are not marked.