CODEWITHSUNDEEP
rtidyversepurrr
Javier Hernando
Javier Hernando

Reputation: 368

Iteration with variables with several time measurements: purrr and map() functions

I have a large database to face with this structure The str of my real database is like this tibble [561 x 128] (S3: tbl_df/tbl/data.frame)

Let's summarise in this dataframe what I need to do

paciente <- c(6430, 6494, 6165, 6278, 6188, 6447, 6207, 6463)
sexo_s1 <-  c("Hombre", "Mujer", "Mujer", "Mujer", "Hombre", "Hombre", "Mujer")
edad_s1 <- c(54, 68, 75, 85, 78, 80, 78, 90)
peso1_v00 <- c(115.2, 85, 98, 87, 85, 78, 84, 98)
cintura1_v00 <- c(115, 125, 110, 114, 120, 121 125, 110)
coltot_v00 <- c(215, 220, 210, 225, 215, 220, 230, 220)
peso1_v66 <- c(110.2, 80, 95, 87, 83, 78, 84, 98)
cintura1_v01 <- c(112, 125, 110, 110, 112, 121 120, 110)
coltot_v01 <- c(210, 210, 205, 215, 215, 210, 230,1 220)
peso1_v01 <- c(110.2, 80, 95, 87, 83, 78, 84, 98)
cintura1_v01 <- c(112, 125, 110, 110, 112, 121 120, 110)
coltot_v01 <- c(210, 210, 205, 215, 215, 210, 230,1 220)

I need to perform several statistical analysis:

Run normality test (shapiro.test and boxplot) across numeric variables (125 out of 128 variables). I am trying to do it whith purrr::map and similars (purrr:map_dfr)

iterative_example<-map_dfr(.x = quos(paciente, sexo_s1, edad_s1, peso1_v00, cintura1_v00, coltot_v00, peso1_v66, cintura1_v66, coltot_v66, peso1_v01, cintura1_v01, coltot_v01), .f = ~ shapiro.test, data = df_example)

error/rlang_error> Argument 1 must be a data frame or a named atomic vector. Backtrace:

purrr::map_dfr(...) dplyr::bind_rows(res, .id = .id)t. If I exchange map_dfr with map I obtain a list which I cannot export or transform into a data.frame

iterative_example<-map(.x = quos(paciente, sexo_s1, edad_s1, peso1_v00, cintura1_v00, coltot_v00, peso1_v66, cintura1_v66, coltot_v66, peso1_v01, cintura1_v01, coltot_v01), .f = ~ shapiro.test, data = df_example)

List of 9:

function(x)

function(x)

function(x)

I cannot export or unnest the list to get the p-value and t results for the time being, but I'll sort it out. However I'd like to get a data.frame.

Similar to this operation I have to run iterative t.test between the variables observed at different times, if there are significative difference between the taken measurements (I've tried the same map function but I get the exact nested list that with shapiro.test For example

t.test(df_example$peso1_v00,  df_example$peso1_v66)
t.test(df_example$cintura1_v00,  df_example$cintura1_v66)

Syntax to recognise the name of variable: "i_variable1_v00" at specific time "v00" and testing with "i_variable1_v66". I've tried: starts_with() but no result

I am not sure how to perform this and export the output

Welch Two Sample t-test

data: df_example$cintura1_v00 and df_example$cintura1_v66 t = -0.051503, df = 10.399, p-value = 0.9599 alternative hypothesis: true difference in means is not equal to 0 95 percent confidence interval: -5.504848 5.254848 sample estimates: mean of x mean of y 117.500 117.625

2 - Create new columns from the values at 0, 6 and 12 months iteratively. I have created the variables but sistematically repeating lines with the variables in the database. An example with a differente variable in my database.

I am looking for sthg to create variables in new columns iteratively between the variables taken at different time moments:

d_peso1_v66: differnce 0- +6 months d_peso1_v01: difference 0 - 12 months

Example with 2 variables without iteration:

df_example<-mutate(df_example, d_peso1_v66 = peso1_v66 - peso1_v00)
df_example<-mutate(df_example, d_coltot_v01 = coltot_v01 - coltot_v00)

d_variable1_v66 = i_variable1_v66 - i_variable1_v00 d_variable1_v01 = i_variable1_v01 - i_variable1_v00

d_variable2_v01 = i_variable2_v66 - i_variable2_v00 d_variable2_v01 = i_variable2_v01 - i_variable2_v00

df_example <-mutate(across(where(is.numeric)),
varname <- paste("varname01", if variable contains "01" )
df_example <- mutate(df, varname = Petal.Width * n)

Not sure if it possible to perform it in one step, or it is necessary create a function and the pass through the database with map function. Sthg like this but making difference (difference_function)

meanofcol <- function(df, col) {
mutate(df, "Mean of {{col}}" := mean({{col}}))
}
meanofcol(iris, Petal.Width)`

And then wiht map function

df_example2 <- map_dfr (.x = df_example, .f = ~ difference_function, data = df_example)

I have been struggling with different approaches that take much more time than what I think it should take if I knew how to write the syntaxis

Upvotes: 0

Views: 182

Answers (1)

Claudio
Claudio

Reputation: 1538

Try this. Note that I deleted duplicate rows in your data.

df <- data.frame(
  paciente = c(6430, 6494, 6165, 6278, 6188, 6447, 6207, 6463),
  sexo_s1 =  c("Hombre", "Mujer", "Mujer", "Mujer", "Hombre", "Hombre", "Mujer", "Hombre"),
  edad_s1 = c(54, 68, 75, 85, 78, 80, 78, 90),
  peso1_v00 = c(115.2, 85, 98, 87, 85, 78, 84, 98),
  cintura1_v00 = c(115, 125, 110, 114, 120, 121, 125, 110),
  coltot_v00 = c(215, 220, 210, 225, 215, 220, 230, 220),
  peso1_v66 = c(110.2, 80, 95, 87, 83, 78, 84, 98),
  cintura1_v01 = c(112, 125, 110, 110, 112, 121, 120, 110),
  peso1_v01 = c(110.2, 80, 95, 87, 83, 78, 84, 98),
  coltot_v01 = c(210, 210, 205, 215, 215, 210, 230, 220))

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(tidyr)
library(broom)

## First you need to convert non-numeric variables recorded as number into factor

df$paciente <- factor(df$paciente)

## Select numeric variables, pivot in long format, analyse

df |> 
  select(where(is.numeric)) |> 
  pivot_longer(everything()) |>
  group_by(name) |> 
  do(tidy(shapiro.test(.$value)))
#> # A tibble: 8 × 4
#> # Groups:   name [8]
#>   name         statistic p.value method                     
#>   <chr>            <dbl>   <dbl> <chr>                      
#> 1 cintura1_v00     0.897  0.270  Shapiro-Wilk normality test
#> 2 cintura1_v01     0.806  0.0330 Shapiro-Wilk normality test
#> 3 coltot_v00       0.958  0.792  Shapiro-Wilk normality test
#> 4 coltot_v01       0.896  0.269  Shapiro-Wilk normality test
#> 5 edad_s1          0.925  0.469  Shapiro-Wilk normality test
#> 6 peso1_v00        0.870  0.149  Shapiro-Wilk normality test
#> 7 peso1_v01        0.905  0.322  Shapiro-Wilk normality test
#> 8 peso1_v66        0.905  0.322  Shapiro-Wilk normality test

## Now select only *_v00 and *_v66, then pivot to longer and separate

df |> 
  select(paciente, matches("_v00|_v01|_v66")) |> 
  pivot_longer(-paciente) |> 
  separate(name, into=c("name", "time"), sep="_") |> 
  pivot_wider(names_from=time, values_from=value) |> 
  group_by(name) |> 
  do(tidy(t.test(.$v00, .$v01)))
#> # A tibble: 3 × 11
#> # Groups:   name [3]
#>   name     estimate estimate1 estimate2 statistic p.value parameter conf.low
#>   <chr>       <dbl>     <dbl>     <dbl>     <dbl>   <dbl>     <dbl>    <dbl>
#> 1 cintura1     2.5      118.      115       0.824   0.424      14.0    -4.01
#> 2 coltot       5        219.      214.      1.42    0.178      13.4    -2.58
#> 3 peso1        1.88      91.3      89.4     0.329   0.747      13.9   -10.4 
#> # … with 3 more variables: conf.high <dbl>, method <chr>, alternative <chr>

Created on 2021-09-16 by the reprex package (v2.0.1)

You can modify the code to get the other comparison (e.g., 00 vs. 66).

Upvotes: 0

Related Questions