Consider utility function to provide summaries of multiple variables

[JosiahParry]

Semi-working idea

library(sfdep)
library(tidyverse)

guerry |> 
  mutate(
    nb = st_contiguity(geometry),
    wt = st_weights(nb)
  ) |> 
  sf::st_drop_geometry() |> 
  summarise(
    avg = mean(crime_pers),
    std_dev = sd(crime_pers),
    I = list(broom::tidy(global_moran_perm(crimea_pers, nb, wt)))
  ) |> 
  tidyr::unnest_wider(I) |> 
  knitr::kable()

summarise_var <- function(x, nb, wt) {
  tibble::tibble(
    avg = mean(x),
    std_dev = sd(x),
    I = list(broom::tidy(global_moran_perm(x, nb, wt)))
  ) |> 
    tidyr::unnest(I)
}

summarize_all_vars <- function(x, nb, wt) {
  lapply(x, summarise_var, nb, wt) |> 
    collapse::rowbind()
}

[DHLocke]

Another approach.. has to be an easier way library(sfdep) library(tidyverse) sums_with_I <- guerry |> select(code_dept, region, department, starts_with('crime')) |> mutate( nb = st_contiguity(geometry), wt = st_weights(nb) ) |> sf::st_drop_geometry() |> summarise(across(where(is.numeric) , .names = "{.col}__{.fn}" # puts a second underscore between variable name and the function (for easy splitting later) , list( mean = mean , sd = sd , I = ~broom::tidy(global_moran_perm(.x, nb, wt))))) # what I like about the approach above is that: # - it is easy to swap mean and sd to median and IQR, min, max, range.. etc.. # - the syntax builds logically on what the across vignette sets up. # what I don't like # - the output is wiiiide: variable, variable_function_1, variable_function_2, next_variable, next_variable_function_1 and more # here's how we might pry out and reshape # but first, make a Not IN function and a list of items to exclude `%nin%` <- Negate(`%in%`) # Trent Reznor function exclude <- c('I.parameter', 'I.method', 'I.alternative') sums_with_I # yuuck: too wide, gross column names with `$` embedded sums_with_I |> t() |> data.frame() |> rownames_to_column() |> tibble() |> separate(rowname, into = c('variable', 'statistic'), sep = '__') |> # double __ makes sep easy while allowing column names with "_" filter(statistic %nin% exclude) |> pivot_wider(id_cols = variable, names_from = statistic, values_from = t.sums_with_I.) |> readr::type_convert() # handsome output, but not a fun process

…

On Wed, Mar 27, 2024 at 12:09 PM Josiah Parry ***@***.***> wrote: Semi-working idea library(sfdep) library(tidyverse) guerry |> mutate( nb = st_contiguity(geometry), wt = st_weights(nb) ) |> sf::st_drop_geometry() |> summarise( avg = mean(crime_pers), std_dev = sd(crime_pers), I = list(broom::tidy(global_moran_perm(crimea_pers, nb, wt))) ) |> tidyr::unnest_wider(I) |> knitr::kable() summarise_var <- function(x, nb, wt) { tibble::tibble( avg = mean(x), std_dev = sd(x), I = list(broom::tidy(global_moran_perm(x, nb, wt))) ) |> tidyr::unnest(I) } summarize_all_vars <- function(x, nb, wt) { lapply(x, summarise_var, nb, wt) |> collapse::rowbind() } — Reply to this email directly, view it on GitHub <#52>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AD23UBQ5HLKXOSY3RHAV4A3Y2LVL5AVCNFSM6AAAAABFLFXFYGVHI2DSMVQWIX3LMV43ASLTON2WKOZSGIYTCMRQGY3DQOA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.***>

[DHLocke]

Maybe a more sophisticated pivot_longer fixes the second piece? https://stackoverflow.com/questions/70700654/pivot-longer-with-names-pattern-and-pairs-of-columns

…

-Dexter

On Wed, Mar 27, 2024 at 3:22 PM Dexter Locke ***@***.***> wrote: Another approach.. has to be an easier way library(sfdep) library(tidyverse) sums_with_I <- guerry |> select(code_dept, region, department, starts_with('crime')) |> mutate( nb = st_contiguity(geometry), wt = st_weights(nb) ) |> sf::st_drop_geometry() |> summarise(across(where(is.numeric) , .names = "{.col}__{.fn}" # puts a second underscore between variable name and the function (for easy splitting later) , list( mean = mean , sd = sd , I = ~broom::tidy(global_moran_perm(.x, nb, wt))))) # what I like about the approach above is that: # - it is easy to swap mean and sd to median and IQR, min, max, range.. etc.. # - the syntax builds logically on what the across vignette sets up. # what I don't like # - the output is wiiiide: variable, variable_function_1, variable_function_2, next_variable, next_variable_function_1 and more # here's how we might pry out and reshape # but first, make a Not IN function and a list of items to exclude `%nin%` <- Negate(`%in%`) # Trent Reznor function exclude <- c('I.parameter', 'I.method', 'I.alternative') sums_with_I # yuuck: too wide, gross column names with `$` embedded sums_with_I |> t() |> data.frame() |> rownames_to_column() |> tibble() |> separate(rowname, into = c('variable', 'statistic'), sep = '__') |> # double __ makes sep easy while allowing column names with "_" filter(statistic %nin% exclude) |> pivot_wider(id_cols = variable, names_from = statistic, values_from = t.sums_with_I.) |> readr::type_convert() # handsome output, but not a fun process On Wed, Mar 27, 2024 at 12:09 PM Josiah Parry ***@***.***> wrote: > Semi-working idea > > library(sfdep) > library(tidyverse) > guerry |> > mutate( > nb = st_contiguity(geometry), > wt = st_weights(nb) > ) |> > sf::st_drop_geometry() |> > summarise( > avg = mean(crime_pers), > std_dev = sd(crime_pers), > I = list(broom::tidy(global_moran_perm(crimea_pers, nb, wt))) > ) |> > tidyr::unnest_wider(I) |> > knitr::kable() > summarise_var <- function(x, nb, wt) { > tibble::tibble( > avg = mean(x), > std_dev = sd(x), > I = list(broom::tidy(global_moran_perm(x, nb, wt))) > ) |> > tidyr::unnest(I) > } > summarize_all_vars <- function(x, nb, wt) { > lapply(x, summarise_var, nb, wt) |> > collapse::rowbind() > } > > — > Reply to this email directly, view it on GitHub > <#52>, or unsubscribe > <https://github.com/notifications/unsubscribe-auth/AD23UBQ5HLKXOSY3RHAV4A3Y2LVL5AVCNFSM6AAAAABFLFXFYGVHI2DSMVQWIX3LMV43ASLTON2WKOZSGIYTCMRQGY3DQOA> > . > You are receiving this because you are subscribed to this thread.Message > ID: ***@***.***> >

[DHLocke]

Here's another abomination guerry |> select(code_dept, region, department, starts_with('crime')) |> mutate( nb = st_contiguity(geometry), wt = st_weights(nb) ) |> sf::st_drop_geometry() |> summarise(across(where(is.numeric) , .names = "{.col}__{.fn}" # puts a second underscore between variable name and the function (for easy splitting later) , list( mean = mean , sd = sd , I = ~broom::tidy(global_moran_perm(.x, nb, wt))))) |> tidylog::mutate_if(is.tibble, ~select(., `__Moran's I` = statistic, `__p` = p.value)) |> unnest() |> tibble::rownames_to_column() |> pivot_longer(-rowname) |> select(-rowname) |> separate(name, sep = '__', into = c('var', 'stat'), convert = TRUE) |> mutate(var = ifelse(var == '', NA, var), stat = str_remove(stat, '[0-9]')) |> tidylog::fill(var, .direction = 'down') |> pivot_wider(id_cols = 'var', names_from = 'stat', values_from = 'value') But it works?

…

On Wed, Mar 27, 2024 at 3:43 PM Dexter Locke ***@***.***> wrote: Maybe a more sophisticated pivot_longer fixes the second piece? https://stackoverflow.com/questions/70700654/pivot-longer-with-names-pattern-and-pairs-of-columns -Dexter On Wed, Mar 27, 2024 at 3:22 PM Dexter Locke ***@***.***> wrote: > Another approach.. has to be an easier way > > library(sfdep) > library(tidyverse) > > sums_with_I <- > guerry |> > select(code_dept, region, department, starts_with('crime')) |> > mutate( > nb = st_contiguity(geometry), > wt = st_weights(nb) > ) |> > sf::st_drop_geometry() |> > summarise(across(where(is.numeric) > , .names = "{.col}__{.fn}" # puts a second underscore > between variable name and the function (for easy splitting later) > , list( mean = mean > , sd = sd > , I = ~broom::tidy(global_moran_perm(.x, nb, > wt))))) > > # what I like about the approach above is that: > # - it is easy to swap mean and sd to median and IQR, min, max, range.. > etc.. > # - the syntax builds logically on what the across vignette sets up. > > # what I don't like > # - the output is wiiiide: variable, variable_function_1, > variable_function_2, next_variable, next_variable_function_1 and more > > # here's how we might pry out and reshape > # but first, make a Not IN function and a list of items to exclude > `%nin%` <- Negate(`%in%`) # Trent Reznor function > exclude <- c('I.parameter', 'I.method', 'I.alternative') > > sums_with_I # yuuck: too wide, gross column names with `$` embedded > > sums_with_I |> > t() |> > data.frame() |> > rownames_to_column() |> > tibble() |> > separate(rowname, into = c('variable', 'statistic'), sep = '__') |> # > double __ makes sep easy while allowing column names with "_" > filter(statistic %nin% exclude) |> > pivot_wider(id_cols = variable, names_from = statistic, values_from = > t.sums_with_I.) |> > readr::type_convert() > > # handsome output, but not a fun process > > > On Wed, Mar 27, 2024 at 12:09 PM Josiah Parry ***@***.***> > wrote: > >> Semi-working idea >> >> library(sfdep) >> library(tidyverse) >> guerry |> >> mutate( >> nb = st_contiguity(geometry), >> wt = st_weights(nb) >> ) |> >> sf::st_drop_geometry() |> >> summarise( >> avg = mean(crime_pers), >> std_dev = sd(crime_pers), >> I = list(broom::tidy(global_moran_perm(crimea_pers, nb, wt))) >> ) |> >> tidyr::unnest_wider(I) |> >> knitr::kable() >> summarise_var <- function(x, nb, wt) { >> tibble::tibble( >> avg = mean(x), >> std_dev = sd(x), >> I = list(broom::tidy(global_moran_perm(x, nb, wt))) >> ) |> >> tidyr::unnest(I) >> } >> summarize_all_vars <- function(x, nb, wt) { >> lapply(x, summarise_var, nb, wt) |> >> collapse::rowbind() >> } >> >> — >> Reply to this email directly, view it on GitHub >> <#52>, or unsubscribe >> <https://github.com/notifications/unsubscribe-auth/AD23UBQ5HLKXOSY3RHAV4A3Y2LVL5AVCNFSM6AAAAABFLFXFYGVHI2DSMVQWIX3LMV43ASLTON2WKOZSGIYTCMRQGY3DQOA> >> . >> You are receiving this because you are subscribed to this thread.Message >> ID: ***@***.***> >> >