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_tocolumn() |> 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 https://github.com/JosiahParry/sfdep/issues/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_tocolumn() |> 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 https://github.com/JosiahParry/sfdep/issues/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_tocolumn() |> 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 https://github.com/JosiahParry/sfdep/issues/52, or unsubscribe https://github.com/notifications/unsubscribe-auth/AD23UBQ5HLKXOSY3RHAV4A3Y2LVL5AVCNFSM6AAAAABFLFXFYGVHI2DSMVQWIX3LMV43ASLTON2WKOZSGIYTCMRQGY3DQOA . You are receiving this because you are subscribed to this thread.Message ID: @.***>