Collapses a CompadreDB object by averaging matrices over levels of one or more grouping variables (e.g. SpeciesAuthor, MatrixPopulation).

For a given study and species, a COM(P)ADRE database may contain multiple matrices, reflecting different combinations of population, year, and/or treatment. Collapsing allows a user to obtain a single 'grand mean matrix' for each group of interest (e.g. MatrixPopulation), and therefore limit pseudoreplication.

All members of a group must have the same matrix dimension (consider adding MatrixDimension as a grouping variable). All members of a group should have the same AnnualPeriodicity and matrix stage class definitions (see cdb_id_stages). Note that Seasonal matrices should not be collapsed using this method (they should be matrix-multiplied rather than averaged).

cdb_collapse(cdb, columns)

## Arguments

cdb A CompadreDB object Vector of grouping variables to collapse over (corresponding to columns within cdb)

## Details

Will give a warning if members of any group do not all share the same AnnualPeriodicity or stage class definitions, or if cdb contains any rows with a MatrixComposite value of "Seasonal".

Prior to collapsing, columns of class 'factor' will be coerced to 'character', and any list-column apart from mat will be removed.

Within a group, rows of a given column are collapsed as follows:

• mat: matrices are averaged using mpm_mean, and stage class definitions are taken from the first group member

• MatrixComposite: return original value if N = 1, else return "Collapsed"

• Lat: re-calculated by averaging Lat column (if available)

• Lon: re-calculated by averaging Lon column (if available)

• SurvivalIssue: re-calculated from the collapsed mat (max(colSums(matU)))

• others: if all elements equal return that unique value, else paste together all unique values separated by "; "

# filter out Seasonal matrices
CompCollapse <- cdb_collapse(CompSub, columns = c("id_stage"))