Returns the mean number of anglers per boat party, taken from an interviews table. This is the multiplier used to expand a count of boats into a count of anglers when the clerk counted boats rather than the people aboard them.
Boats move, so a count of anglers aboard is often less reliable than a count of hulls. Counting boats and expanding by the interviewed party size trades an unreliable field count for a measured one, at the cost of assuming the interviewed parties are representative of the boats that were counted.
Arguments
- interviews
A data frame of interviews, one row per party.
- n_anglers
Tidy selector for the numeric party-size column.
- angler_type
Optional tidy selector for the column recording whether a party fished from a boat or the bank. When supplied, only boat parties are used.
- boat_value
Value of
angler_typemarking a boat party. Defaults to"boat". Ignored whenangler_typeisNULL.- by
Optional tidy selector for one or more grouping columns. When supplied, a mean is returned for each group rather than one overall value.
Value
When by is NULL, a single numeric value. Otherwise a tibble with
the grouping columns and a mean_party_size column.
Either way the return carries a "se" attribute holding the standard error
of the mean (sd / sqrt(n) over parties), one value per group for the by
form. derive_angler_count() reads it, so the sampling error of the
multiplier reaches the effort standard error without being passed by hand.
The standard error is an attribute rather than a column so that the scalar
return stays usable directly as a multiplier, and so the by form keeps
exactly one numeric column and remains valid as a party_size lookup.
For the by form the attribute is named by the group key, and
derive_angler_count() addresses it by name. Attributes do not follow the
rows they describe through a dplyr reordering, so a positional attribute
would go stale the moment the lookup were sorted — attributing each
stratum's standard error to a different stratum while the means, which join
by key, stayed correct. A by-form lookup whose "se" attribute has no
names is refused rather than matched by row order.
A group with a single party has no estimable standard error and gets
NA_real_, which propagates to an NA effort standard error rather than
being quietly treated as zero uncertainty.
Details
Each row of interviews is assumed to be one party. A table carrying several
rows per party — one per species, say — will weight larger parties more than
once; reduce it to one row per party first.
Supply by when party size differs across the survey. Weekend parties are
commonly larger than weekday parties, and a single season-wide mean applied to
both then moves effort in opposite directions in the two strata.
See also
derive_angler_count(), prep_counts_boat_party()
Other "Survey Design":
add_catch(),
add_counts(),
add_interviews(),
add_lengths(),
add_sections(),
as_creel_svydesign(),
as_hybrid_svydesign(),
compute_angler_effort(),
compute_effort(),
creel_design(),
creel_schema(),
creel_vocabulary(),
derive_angler_count(),
est_effort_camera(),
impute_camera_counts(),
prep_counts_boat_party(),
prep_counts_daily_effort(),
prep_interview_catch(),
prep_interviews_trips(),
validate_creel_schema()
Examples
interviews <- data.frame(
day_type = c("weekday", "weekday", "weekend", "weekend"),
type = c("boat", "bank", "boat", "boat"),
n_anglers = c(2, 1, 3, 4)
)
# Overall, boat parties only
mean_party_size(interviews, n_anglers, angler_type = type)
#> [1] 3
#> attr(,"se")
#> [1] 0.5773503
# By stratum
mean_party_size(interviews, n_anglers, angler_type = type, by = day_type)
#> # A tibble: 2 × 2
#> day_type mean_party_size
#> <chr> <dbl>
#> 1 weekday 2
#> 2 weekend 3.5