1 Directory commands

• getwd() tells you what is the path of the R project (.Rproj)

getwd()

## [1] "D:/Dropboxvhir/Nuevo Equipo VHIR10 Dropbox/Estela Ariza Moral/UEB-compartida/Aplicacions i Programari/R_Manual"

• setwd() allows to set/change the directory for the R project (.Rproj)

# This example sets the directory in Desktop folder, R project is saved in "e:/user.name/Desktop/my_project.Rproj"
setwd("e:/user.name/Desktop")

• dir() extracts the file names, by default extract the file names of the current working directory of the R project

# same as dir(getwd()):
dir()

## [1] "data"                            "diabetes.xls"                   
## [3] "my_work_space.RData"             "R_Manual-DataManipulation.html" 
## [5] "R_Manual-DataManipulation.Rmd"   "R_Manual-DataManipulation.Rproj"
## [7] "Rcode"

Get the file names of another working directory:

path <- "C:/user.name/directory/file/"

dir(path)

## character(0)

Print the objects in environment:

print(ls())

## [1] "path"

Remove objects from environment:

# removes all objects from memory
rm(list=ls())

print(ls())
# remove specific objects
rm(dat, path)

2 Good to know

• Packages require() and library() both load a R package, but they have a main difference:
  • require() will output a warning if a package is not installed and then continue to execute the code. It is usually only used if you are loading packages inside a function so that the function will continue to execute even if a package doesn’t exist.
  • library() will output an error and stop the execution of the code.
• To install old versions of packages, try these options in a new R Script:

# option 1

remove.packages("knitr") # remove the actual version
require(remotes)
remotes::install_version(package ="knitr", version ="1.43", repos=getOption("repos"))


# option 2

remove.packages("kableExtra") # remove the actual version
require(utils)
packageurl <- "https://cran.r-project.org/src/contrib/Archive/kableExtra/kableExtra_1.3.4.tar.gz"
utils::install.packages(pkgs = packageurl, repos = NULL,type = "source")

• p_load() allows to install and load a R package, but first install and load package pacman.

  • Example: p_load(ggplot2, dplyr)

• identical() returns boolean TRUE/FALSE if two R objects (data frames, vectors, matrices, …) are the same.

  • Example: identical(dat1, dat2)

• To access for more information for:

  • For a given function of a R package, in the console type “?func”.
    • Example: ?identical
  • For a package:
    • Example: help(package=“dplyr”)

• Load package magrittr to use operator %<>% in R.

• apply(), lapply(), sapply():

  • apply(): lets us apply a function to the rows or columns of a matrix or data frame.
    • Example: apply( sample_matrix, 2, mean) -> use apply() function across column to find mean of each column
  • lapply(): lets us apply a function to list objects.
    • Example: lapply(names, toupper) -> all strings in list names are in capital letters
  • sapply(): lets us apply a function to lists, vectors, data frames and returns an array or matrix object.
    • Example: sapply(sample_data, max) -> get the maximum from columns of data set sample_data

• Function filter() + combined with operator ! -> be careful. For example, assume you have a factor variable [cat] that takes values “No”, “Yes” and some are NA. If you do dat %>% filter(cat!=“No”) it will exclude those rows where NA in [cat], but maybe you are still interested in keeping those cases even if they have [cat]=NA. In that case, do dat %>% filter(cat!=“No” | is.na(cat)).

  • Example: dat %>% filter(grupo_1!=“0” | is.na(grupo_1))

• Use a particular function from a package. This is useful when you load two packages that have a function with the same name.

  • Example: dplyr::select() -> needs to be specified if in the same project you load package dplyr and MASS (MASS::select()). Both packages have a select() function.

• Load built-in dataset into the environment

  • Example: data(iris)

• Difference between paste() and paste0():

  • paste0(): allows concatenating strings but by default does not leave a blank space in between strings. Example: paste0(“ID”, seq(10))
  • paste(): allows concatenating strings and by default leaves a blank space in between strings. Example: paste(“ID”, seq(10))

3 Read data

p_load(readxl)
dat <- read_excel("diabetes.xls") # read_excel("diabetes.xlsx")
head(dat)

class(dat)

## [1] "tbl_df"     "tbl"        "data.frame"

p_load(rio)
# form rio package -> same as rio::import("diabetes.xls")
dat <- import("diabetes.xls")
class(dat)

## [1] "data.frame"

# export 
export("diabetes.xls")

# Saving on object in RData format
save(dat, file = "my_data.RData")
# Save multiple objects
save(data1, data2, file = "my_data.RData")
# To load the data again
load("my_data.RData")

# export
save.image(file = "my_work_space.RData")
# import
load("my_work_space.RData")

# export
saveRDS(dat, file="my_data.rds")
# import
readRDS(file="my_data.rds")

4 Basic data frame operations

From now on, we use package dplyr.

p_load(dplyr, magrittr, forcats)

Display structure of data frame:

str(dat)

## 'data.frame':    149 obs. of  11 variables:
##  $ numpacie: num  1 2 3 4 5 6 7 8 9 10 ...
##  $ mort    : chr  "Vivo" "Vivo" "Vivo" "Vivo" ...
##  $ tempsviu: num  12.4 12.4 9.6 7.2 14.1 14.1 12.4 14.2 12.4 14.5 ...
##  $ edat    : num  44 49 49 47 43 47 50 36 50 49 ...
##  $ bmi     : num  34.2 32.6 22 37.9 42.2 33.1 36.5 38.5 41.5 34.1 ...
##  $ edatdiag: num  41 48 35 45 42 44 48 33 47 45 ...
##  $ tabac   : chr  "No fumador" "Fumador" "Fumador" "No fumador" ...
##  $ sbp     : num  132 130 108 128 142 156 140 144 134 102 ...
##  $ dbp     : num  96 72 58 76 80 94 86 88 78 68 ...
##  $ ecg     : chr  "Normal" "Normal" "Normal" "Frontera" ...
##  $ chd     : chr  "No" "No" "Si" "Si" ...

Display most relevant statistics of all variables in data frame:

summary(dat)

##     numpacie          mort              tempsviu          edat      
##  Min.   :  1.00   Length:149         Min.   : 0.00   Min.   :31.00  
##  1st Qu.: 38.00   Class :character   1st Qu.: 7.30   1st Qu.:43.00  
##  Median : 75.00   Mode  :character   Median :11.60   Median :50.00  
##  Mean   : 75.01                      Mean   :10.52   Mean   :52.17  
##  3rd Qu.:112.00                      3rd Qu.:13.90   3rd Qu.:60.00  
##  Max.   :149.00                      Max.   :16.90   Max.   :86.00  
##       bmi           edatdiag        tabac                sbp       
##  Min.   :18.20   Min.   :26.00   Length:149         Min.   : 98.0  
##  1st Qu.:26.60   1st Qu.:38.00   Class :character   1st Qu.:124.0  
##  Median :31.20   Median :45.00   Mode  :character   Median :138.0  
##  Mean   :31.78   Mean   :45.99                      Mean   :139.1  
##  3rd Qu.:35.20   3rd Qu.:53.00                      3rd Qu.:152.0  
##  Max.   :59.70   Max.   :81.00                      Max.   :222.0  
##       dbp             ecg                chd           
##  Min.   : 58.00   Length:149         Length:149        
##  1st Qu.: 74.00   Class :character   Class :character  
##  Median : 80.00   Mode  :character   Mode  :character  
##  Mean   : 90.04                                        
##  3rd Qu.: 88.00                                        
##  Max.   :862.00

# for a particular column
summary(dat$numpacie)

##    Min. 1st Qu.  Median    Mean 3rd Qu.    Max. 
##    1.00   38.00   75.00   75.01  112.00  149.00

Reset row index:

# useful when you have previously sliced (filtered) the data frame using dat[rows_filtered, ]
rownames(dat) <- NULL

Generate new column that takes the values of the row indices:

# option 1
dat %>% tibble::rownames_to_column(var="Old_RowIndex")

# option 2
dat %<>% mutate(Old_RowIndex=row_number())

Create empty data frame:

# data.frame(col1_name=character(n), col2_name=character(n), col3_name=character(n))
data.frame(col1=character(2), col2=integer(2), col3=numeric(2)) # n=2 -> empty data frame with two rows

5 Row-wise operations

# option 1
# dat %>% filter(numpacie %in% names(which(table(dat$numpacie) > 1))) %>% View() -> same but it previous the operation in the View window
dat %>% filter(numpacie %in% names(which(table(dat$numpacie) > 1)))

# To verify that filter is correct:
dat %>% filter(numpacie=="135")# indeed, numpacie 135 has two records

# option 2
dat %>% arrange(numpacie) %>% group_by(numpacie) %>% filter(n()>1)

# dat %>% filter(numpacie %in% names(which(table(dat$numpacie) > 1))) %>% select(numpacie, mort, edat, bmi) %>% View() -> same but it previous the operation in the View window
dat %>% filter(numpacie %in% names(which(table(dat$numpacie) > 1))) %>% select(numpacie, mort, edat, bmi)

dat %>% group_by(numpacie) %>% filter(n_distinct(id)>1)

# To verify that filter is correct:
dat %>% filter(numpacie=="135")# indeed, numpacie 135 is assigned to two ids

dat %>% group_by(id) %>% filter(n_distinct(numpacie)>1)

# option 1
dat <- dat[apply(dat, 1, function(row) all(is.na(row))), ]

# option 2
dat <- dat[rowSums(is.na(dat))==ncol(dat), ]

dat <- dat[ !(is.na(dat$id) | rowSums(is.na(dat))==ncol(dat)), ]

# find row index where "PAT_ID" appears in any column
row_idx <- which(apply(dat, 1, function(row) any(row=="PAT_ID")))[1]
# remove rows before row index where "PAT_ID" appears in any column
dat <- dat[(row_idx):nrow(dat), ]

# option 1
dat[-which(dat$mort == "Vivo" & dat$tabac == "No fumador"),]

# option 2 -> row index is the count of rows that match condition
dat %>% filter(!(mort == "Vivo" & dat$tabac == "No fumador"))

identical(dat %>% filter(!(mort == "Vivo" & dat$tabac == "No fumador")), dat[-which(dat$mort == "Vivo" & dat$tabac == "No fumador"),])

## [1] FALSE

# option 1
dat[which(dat$tabac == "No fumador"),]

# option 2
dat %>% filter(tabac == "No fumador")

# To check that filter is correct:
dat %>% filter(tabac == "No fumador") %>% nrow()# is 57

## [1] 57

table(dat$tabac)#count is 57 for 'No fumador'

## 
## Ex fumador    Fumador No fumador 
##         41         51         57

# option 1
dat[which(dat$mort == "Vivo" & dat$tabac == "No fumador"),]

# option 2
dat %>% filter(mort == "Vivo" & dat$tabac == "No fumador")

dat %>% filter(grepl("veces|día", fruta_freq)) %>% select(numpacie, fruta_freq)

# same as: dat %>% filter(if_any(everything(), ~. == "No Info"))
dat %>% filter(if_any(everything(), ~. %in% "No Info"))

# alternative: dat %>% filter(if_all(everything(), ~. == "No Info"))
dat %>% filter(if_all(everything(), ~. %in% "No Info"))

dat %>% filter(if_any(.cols=-c(numpacie, edat), ~. %in% "No Info"))

dat %>% filter(if_any(.cols=where(is.numeric),~.>500))

dat %>% filter(numpacie %in% c("8", "19", "20"))

dat %>% filter(!numpacie %in% c("8", "19"))

dat %>% filter(!numpacie %in% c("8", "19") & edat > 40)

dat %>% filter(!(edat>40) & !numpacie %in% c("8", "19"))

dat1 %>% group_by(id) %>% filter(row_number() == max(which(complete.cases(vivo)))) %>% ungroup() %>% as.data.frame()

data <- data.frame(
  ID = c("01", "01", "01", "02", "02"),
  X = c(NA, 1, 1, NA, NA),
  Y = c(NA, 200, NA, 1, NA),
  Z = c(NA, NA, NA, 1, 1)
)

data %>% group_by(ID) %>% summarise_all(function(x) ifelse(all(is.na(x)), NA, na.omit(x)[1]))

6 Column-wise operations

Select specific columns:

# alternative: dat %>% select(mort, bmi))
dat %>% select(c(mort, bmi))

Select first five columns:

dat %>% select(c(head(names(.),5)))

Unselect column(s):

# alternative: dat %>% select(-mort, -bmi))
dat %>% select(-c(mort, bmi))

Uselect last five columns:

dat %>% select(-c(tail(names(.),5)))

dat %>% select(where(is.numeric)) %>% select_if(~all(.>500)) %>% ncol()

## [1] 0

dat %>% select(where(is.numeric)) %>% select_if(~any(.>500))

# verify that filter is correct -> there are two rows with values greater than 500 in column dbp
dat %>% select(where(is.numeric)) %>% filter(dbp >500)

dat %>% select(where(~any(. %in% c("Sense especificar", "N/A"))))

dat %<>% select(-where(~all(is.na(.))))

library(dplyr)
# Remove  variables with complete in the name
dat %>% select(-contains("complete"))
# Select  variables with complete in the name
dat %>% select(contains("complete"))

dat %>% select(numpacie, edat, everything())

class(dat$mort)

## [1] "character"

# option 1
dat$mort <- as.factor(dat$mort)
dat$tabac <- as.factor(dat$tabac)
dat$chd <- as.factor(dat$chd)
dat$ecg <- as.factor(dat$ecg)

# option 2, instead of as.factor, we could also do as.numeric, as.character
dat %<>% mutate_at(vars(c(mort, tabac, chd, ecg)), as.factor)

dat %<>% mutate_all(as.character)

int64_columns <- sapply(df, function(x) any(class(x) == "integer64"))

# Use dplyr to convert the identified columns to numeric
df <- df %>%
  mutate(across(all_of(names(df)[int64_columns]), as.numeric))

var_date <- names(dat %>% select(grep("fecha",names(.),value = T)))
dat %<>%  mutate_at(all_of(var_date), list(~as.Date(., "%d-%m-%Y")),.keep = all)

var_fac <- names(dat %>% select(mort, starts_with("grupo_")))
dat %<>% mutate_at(all_of(var_fac), as.factor)

# option 1, can also do: names(dat)
colnames(dat)

##  [1] "numpacie"   "mort"       "tempsviu"   "edat"       "bmi"       
##  [6] "edatdiag"   "tabac"      "sbp"        "dbp"        "ecg"       
## [11] "chd"        "id"         "fruta_freq"

# option 2
dat %>% names()

##  [1] "numpacie"   "mort"       "tempsviu"   "edat"       "bmi"       
##  [6] "edatdiag"   "tabac"      "sbp"        "dbp"        "ecg"       
## [11] "chd"        "id"         "fruta_freq"

# option 1
dat %>% select(tempsviu:ecg) %>% names()

## [1] "tempsviu" "edat"     "bmi"      "edatdiag" "tabac"    "sbp"      "dbp"     
## [8] "ecg"

# option 2
names(dat %>% select(tempsviu:ecg))

## [1] "tempsviu" "edat"     "bmi"      "edatdiag" "tabac"    "sbp"      "dbp"     
## [8] "ecg"

"num" %in% colnames(dat)

## [1] FALSE

"numpacie" %in% colnames(dat)

## [1] TRUE

any(grepl("num", colnames(dat)))

## [1] TRUE

colnames(dat)[grepl("num", colnames(dat))]

## [1] "numpacie"

# option 1: dat only keeps columns that start with "grupo_" and renames them
dat %<>%
    select(starts_with("grupo_")) %>%
    setNames(sub("grupo_(\\d+)_.*", "grupo_\\1", colnames(.)))

# option 2: keeps all columns and renames columns that start with "grupo_"
grupo_columns <- colnames(dat)[startsWith(colnames(dat), "grupo_")]

# Rename the selected columns
colnames(dat)[startsWith(colnames(dat), "grupo_")] <- sub("grupo_(\\d+)_.*", "grupo_\\1", grupo_columns)

p_load(janitor)
dat <- janitor::clean_names(dat)

# option 1 -> by referencing column name
colnames(dat)[colnames(dat) == "numpacie"] <- "id_paciente" 

colnames(dat)[colnames(dat) %in% c("numpacie", "tempsviu")] <- c("id_paciente", "temps_viu")

# option 2 -> not recommended, "numpacie" might not always be the first column
colnames(dat)[1] <- "id_paciente" 

colnames(dat)[c(1,3)] <- c("id_paciente", "temps_viu")

# option 3
dat %>% colnames()

##  [1] "numpacie"   "mort"       "tempsviu"   "edat"       "bmi"       
##  [6] "edatdiag"   "tabac"      "sbp"        "dbp"        "ecg"       
## [11] "chd"        "id"         "fruta_freq"

dat %>% rename("id_paciente"="numpacie", "temps_viu"="tempsviu") %>% colnames()

##  [1] "id_paciente" "mort"        "temps_viu"   "edat"        "bmi"        
##  [6] "edatdiag"    "tabac"       "sbp"         "dbp"         "ecg"        
## [11] "chd"         "id"          "fruta_freq"

str_suffix = "A"
dat %>% mutate(!!glue::glue("grup_{str_suffix}") = "A")

str_suffix = "hosp"
dat %>% group_by(id) %>% dplyr::summarize(!!glue::glue("num_{str_suffix}") := n())

# extract first row
col_names <- as.character(dat[1,])
# remove first row from dat
dat <- dat[-1, ]
# set first row as column names
colnames(dat) <- col_names

dat %>% rename_with(~paste0(., "_dataset1"))

# based on a parameter
time_str <- "basal"
dat %>% rename_with(~paste0(., glue::glue("_{time_str}")))

dat %>% rename_at(vars(-numpacie), ~paste0(., "_dataset1"))

# based on a parameter
time_str <- "basal"
dat %>% rename_at(vars(-numpacie), ~paste0(., glue::glue("_{time_str}")))

dat %>% colnames()

##  [1] "numpacie"   "mort"       "tempsviu"   "edat"       "bmi"       
##  [6] "edatdiag"   "tabac"      "sbp"        "dbp"        "ecg"       
## [11] "chd"        "id"         "fruta_freq"

# first we add suffix to all column names
dat %<>% rename_with(~paste0(., "_", seq_along(.)))
dat %>% colnames()

##  [1] "numpacie_1"    "mort_2"        "tempsviu_3"    "edat_4"       
##  [5] "bmi_5"         "edatdiag_6"    "tabac_7"       "sbp_8"        
##  [9] "dbp_9"         "ecg_10"        "chd_11"        "id_12"        
## [13] "fruta_freq_13"

p_load(stringr)# for str_remove()
dat %>% colnames()

##  [1] "numpacie_1"    "mort_2"        "tempsviu_3"    "edat_4"       
##  [5] "bmi_5"         "edatdiag_6"    "tabac_7"       "sbp_8"        
##  [9] "dbp_9"         "ecg_10"        "chd_11"        "id_12"        
## [13] "fruta_freq_13"

# then we remove "_digit"
dat %<>% rename_with(~str_remove(.,"_\\d+$"), everything())
dat %>% colnames()

##  [1] "numpacie"   "mort"       "tempsviu"   "edat"       "bmi"       
##  [6] "edatdiag"   "tabac"      "sbp"        "dbp"        "ecg"       
## [11] "chd"        "id"         "fruta_freq"

dat %>% rename_with(~paste0("var_",.))

# based on a parameter
time_str <- "basal"
dat %>% rename_with(~paste0(glue::glue("_{time_str}", .)))

p_load(Hmisc)
Hmisc::label(dat)

##   numpacie       mort   tempsviu       edat        bmi   edatdiag      tabac 
##         ""         ""         ""         ""         ""         ""         "" 
##        sbp        dbp        ecg        chd         id fruta_freq 
##         ""         ""         ""         ""         ""         ""

# option 1 -> by referencing column name
Hmisc::label(dat[,"numpacie"]) <- "ID Paciente" 

Hmisc::label(dat,self = F)[names(dat) %in% c("numpacie", "tempsviu")] <- c("ID Paciente", "Tiempo Vivo")

# option 2 -> not recommended, "numpacie" might not always be the first column
Hmisc::label(dat)[1] <- "ID" 

Hmisc::label(dat,self = F)[c(1,3)] <- c("ID Paciente", "Tiempo Vivo")

# labels_cols <- sapply(dat1, Hmisc::label) if you want to have same labels as dat1 in dat
labels_cols <- c("ID", "muerte", "tiempo vivo", "edad", "BMI", "edad diag", "tabaco", "SBP", "DBP", "ECG", "CHD")
Hmisc::label(dat) <- as.list(labels_cols)

  id, time, alive_v1, alive_v2, alive_v3,
  ID1, t0, 1, NA, NA, 
  ID1, t1, NA, 1, NA,
  ID1, t2, NA, NA, 0

  id, time, alive,
  ID1, t0, 1,
  ID1, t1, 1,
  ID1, t2, 0

dat %<>% mutate( alive = coalesce(alive_v1, alive_v2, alive_v3)) %>% dplyr::select(-c(alive_v1, alive_v2, alive_v3))

# option 1
dat %<>% mutate(fruta_freq=factor(case_when(edat<30 ~ "Nunca", edat <=30 & edat < 40 ~ "A veces", 
                                    edat>=40 & edat<50 ~ "Bastantes veces", 
                                    edat>=50 & edat<60 ~ "Cada dos días", 
                                    edat >= 60 & edat < 70 ~ "Varias al día", 
                                    edat >= 70 ~ "Siempre"), levels=c("Nunca", "A veces", "Bastantes veces", "Cada dos días", "Varias al día", "Siempre")))

# option 2
dat$fruta_freq <- case_when(dat$edat<30 ~ "Nunca", dat$edat <=30 & dat$edat < 40 ~ "A veces", 
          dat$edat>=40 & dat$edat<50 ~ "Bastantes veces", dat$edat>=50 & dat$edat<60 ~ "Cada dos días", dat$edat >= 60 & dat$edat < 70 ~ "Varias al día", dat$edat >= 70 ~ "Siempre")

p_load(stringr)
dat %<>% mutate(fumador=factor(case_when(str_detect(tolower(tabac), "^no") | str_detect(tolower(tabac), "^ex") ~ "No",
                                         str_detect(tolower(tabac), "^fumador") ~ "Si"), 
                               levels=c("No", "Si")))

dat %>% mutate(start_date=sample(seq(as.Date('1999/01/01'), as.Date('2000/01/01'), by="day"), 149), 
               end_date=sample(seq(as.Date('2000/01/02'), as.Date('2000/12/31'), by="day"), 149), 
               duration=as.numeric(difftime(end_date, start_date)))

# na.rm=TRUE -> do the counting even if there is a NA in one of the "V" columns
as.data.frame(matrix(sample(c(0, 1, NA), 149*20, replace = TRUE), ncol=20)) %>% mutate(id=paste0("ID", seq(1:149))) %>% select(id, everything()) %>% mutate(num_illnesses=rowSums(select(., c(starts_with("V"))), na.rm=TRUE)) %>% head()

dat <- replace(dat, dat == "Sin especificar", NA)
# could also do: replace(dat, dat == "Sin especificar" | "N/A", NA)

# option 1
dat$edat[dat$numpacie==1] <- 10

# option 2
## alternative: dat[which(dat$numpacie==1),]$edat
dat[which(dat$numpacie==1),"edat"] <- 10

# option 3
## before
dat %>% filter(numpacie==1) %>% select(numpacie, edat)

## after
dat %>% mutate(edat = case_when(numpacie==1 ~ 10, numpacie!=1 ~ edat))  %>% filter(numpacie==1) %>% select(numpacie, edat)

# option 1
## before
dat$edat[dat$tabac=="No fumador" & dat$ecg=="Normal"]

##  [1] 10 47 49 54 40 38 44 45 35 44 47 78 71 66 42 67 86 42 60 81 60 45 61 82 50
## [26] 71 69 48 58 52 52 59 49 60 61 54 52 75

## after
dat$edat[dat$tabac=="No fumador" & dat$ecg=="Normal"] <- 10
dat$edat[dat$tabac=="No fumador" & dat$ecg=="Normal"]

##  [1] 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
## [26] 10 10 10 10 10 10 10 10 10 10 10 10 10

# option 3
## before
dat %>% filter(tabac=="No fumador" & ecg=="Normal") %>% select(numpacie, edat)

## after
dat %>% mutate(edat = case_when(tabac=="No fumador" & ecg=="Normal" ~ 10, 
                                !(tabac=="No fumador" & ecg=="Normal") | (is.na(tabac) | is.na(ecg)) ~ edat))  %>% 
  filter(tabac=="No fumador" & ecg=="Normal") %>% select(numpacie, edat)

  id, time, gender
  01, 0, male
  01, 1, NA
  01, 2, NA

  id, time, gender
  01, 0, male
  01, 1, male
  01, 2, male

# option 1
dat %<>% group_by(numpacie) %>% mutate(gender=gender[!is.na(gender)]) %>% ungroup() %>% as.data.frame()

# option 2 -> useful if you want to do it for more than one variable at the same time
dat %>% group_by(numpacie) %>% fill(c("gender"), .direction = "down") %>% ungroup() %>% as.data.frame()

table(dat$mort)

## 
##  Muerto No Info    Vivo 
##      24       5     120

dat %>%  mutate(mort = plyr::revalue(mort, c("Vivo" = "alive", "Muerto" = "dead"))) %>% pull(mort) %>% table()

## .
##   alive    dead No Info 
##     120      24       5

# option 1
dat$x1 <- fct_collapse(dat$x1, "1"=levels(x1)[-1])

# option 2
dat %>% mutate(x1=fct_collapse(x1, "1"=levels(x1)[-1]))

# option 1
dat$tabac <- fct_collapse(dat$tabac, "No fumador"= c("Ex fumador", "No fumador"))

## could also just simply create new variable instead of replacing original variable
dat$fumador <- fct_collapse(dat$tabac, "No fumador"= c("Ex fumador", "No fumador"))

# option 2
dat %>% mutate(tabac=fct_collapse(tabac, "No fumador"= c("Ex fumador", "No fumador")))

## could also just simply create new variable instead of replacing original variable
dat %>% mutate(fumador=fct_collapse(tabac, "No fumador"= c("Ex fumador", "No fumador")))

# Change all values of  factors variables to NA or the same value 

for (i in 1:ncol(dat)) {
  if (is.factor(dat[[i]])) {
    levels(dat[[i]])[levels(dat[[i]]) %in% c("Unknown")] <- NA
  }
}

dat$hour <- sample(c(round(runif(20, min=0, max=23.59), 2), "N/A", "09/09/9999"), 149, replace=TRUE)

# first check if 9999 is in column
any(grepl("9999", dat$hour))
# check if string contains letters such as N, A from N/A
any(grepl("[a-zA-Z]+", dat$hour))

# now replace any values that are not digit with NA
dat$hour <- ifelse(grepl("\\d", dat$hour), dat$hour, NA)

7 Row-wise and column-wise operations

• Select [numpacie] column and columns with column names that start with “grupo_”
• For columns with column names that start with “grupo_”, filter rows that have values different than “0”.

For columns with column names that start with “grupo_”, filter rows that have values different than “0” in all of the columns that start with “grupo_”:

dat %>% select(numpacie, starts_with("grupo_")) %>% filter_at(vars(starts_with("grupo_")), all_vars(.!="0"))

For columns with column names that start with “grupo_”, filter rows that have at least a value different than “0” in one of the columns that start with “grupo_”:

dat %>% select(numpacie, starts_with("grupo_")) %>% filter_at(vars(starts_with("grupo_")), any_vars(.!="0"))

dat_grups <- dat %>% group_by(id)
ls_grups <- group_split(dat_grups)
# access first group
ls_grups[[1]]

res <- ls_grups %>% map(~ filter(.x, id_cas == "1141959"))
res[!sapply(res, function(df) nrow(df) == 0)]

8 Compare data frames

Check which IDs in dataset 1 are not present in dataset 2, assuming that ID codification is the same in both data sets:

setdiff(dat1$id,dat2$id)

dat1 %>% filter(id %in% dat2$id) %>% nrow()

# Calculate counts of IDs in each dataset
counts_dat1 <- table(dat1$id_cas)
counts_dat2 <- table(dat2$id_cas)

# Find all unique IDs across both datasets
all_ids <- union(names(counts_dat1), names(counts_dat2))

# Initialize a data frame to store the results
result_df <- data.frame(ID = all_ids, Count_in_dat1 = 0, Count_in_dat2 = 0)

# Fill in the counts for each dataset
result_df$Count_in_dat1[result_df$ID %in% names(counts_dat1)] <- counts_dat1[result_df$ID %in% names(counts_dat1)]
result_df$Count_in_dat2[result_df$ID %in% names(counts_dat2)] <- counts_dat2[result_df$ID %in% names(counts_dat2)]

intersect(names(df1), names(df2))

9 Join data frames

Can join data frames by using merge() or join() functions from dplyr package. In tidyverse, a “merge” is called a “join” which is the term more commonly used in relational databases.. This manual works with join():

This is are most common types of joins:

• inner_join() corresponds to merge(X, Y) – only retain rows that match between X and Y.

• left_join() corresponds to merge(X, Y, all.x = T) – retain all rows from X, whether or not they match in Y.

• right_join() corresponds to merge(X, Y, all.y = T) – retain all rows from Y, whether or not they match in X.

• full_join() corresponds to merge(X, Y, all = T) – retain all rows from both X and Y, whether or not they match.

# option 1
dat <- full_join(dat1, dat2)

# option 2 using pipe
dat <- dat1 %>% full_join(dat2)

10 Missing data

dat2 <- as.data.frame(matrix(sample(c(0, 1, NA), 149*20, replace = TRUE), ncol=20)) %>% mutate(id=paste0("ID", seq(1:149))) %>% select(id, everything())
head(dat2)

Check if data frame has at least one NA:

any(is.na(dat2))

## [1] TRUE

Count number of missings for each column:

# option 1
sapply(dat2, function(x) sum(is.na(x)))

##  id  V1  V2  V3  V4  V5  V6  V7  V8  V9 V10 V11 V12 V13 V14 V15 V16 V17 V18 V19 
##   0  51  56  44  55  54  49  55  51  49  49  50  39  50  41  48  38  57  44  56 
## V20 
##  58

# option 2
colSums(is.na(dat2))

##  id  V1  V2  V3  V4  V5  V6  V7  V8  V9 V10 V11 V12 V13 V14 V15 V16 V17 V18 V19 
##   0  51  56  44  55  54  49  55  51  49  49  50  39  50  41  48  38  57  44  56 
## V20 
##  58

Count number of missings for each row:

rowSums(is.na(dat2))

##   [1]  5  2  6  2  8  6  7  6  8  5  5  7  9  3  5  8  7  9  4  3  5 11  9  4  6
##  [26] 10  4 10  6  7  8  4 10  5  8  9  3 11 10  6  5  2  2 10  7  5  8  8 13  7
##  [51]  8  7  6  6  7  4  6  4 10 11  6  7  7  6  5  6  7  5  7  7  9  3  9  6 11
##  [76]  9  6  4  4  5 12  7  6  6  8  5  4  4  9  6  6  6  5  7 11  5  9  7  7 12
## [101]  6  4  6  6  9  5  6  7  7 12  8  5  9  7  3  7  8  5  6  9  5  6  7  6  9
## [126]  7  8  5  7 10  6  7  8  6  8  5  5  9  6  9  6  5  7  7  3  4  8  7  5

# alternative -> rowSums(is.na(dat2) | dat2 == "" | dat2 == " ")

# create column that counts NAs per row
# alternative -> dat2 %>% mutate(row_count_na=rowSums(is.na(select(., -id))))
dat2 %>% mutate(row_count_na=rowSums(is.na(select(., everything()))))

11 Generate matrix

Slicing:

• Select a row: mat[2, ] -> select row 2
• Select a column: mat[,2] -> select second column
• Select an element: mat[2,2] -> select element in row 2 and column 2

# option 1
matrix(c(1,1, 2,2), nrow=2, ncol=2)

##      [,1] [,2]
## [1,]    1    2
## [2,]    1    2

as.data.frame(matrix(c(1,1, 2,2), nrow=2, ncol=2))

# option 2 -> matrix(c(col1_vector, col2_vector), 2)
matrix(c(c(1,1), c(2,2)), ncol=2)

##      [,1] [,2]
## [1,]    1    2
## [2,]    1    2

# option 3 -> column wise: cbind(col1_vector, col2_vector)
cbind(c(1,1), c(2,2))

##      [,1] [,2]
## [1,]    1    2
## [2,]    1    2

# option 4 -> row wise: rbind(row1_vector, row2_vector)
rbind(c(1,2), c(1,2))

##      [,1] [,2]
## [1,]    1    2
## [2,]    1    2

12 Plots

var_out <- "bsa"
var_comp <-  "contraste"

ggplot(dat, aes(!!sym(var_out))) + geom_bar(aes(fill = !!sym(var_comp)), position = "dodge")

# first plot is the density function of var_num grouped by var_fact
plot1 <- ggplot(dat, aes(x = var_num, colour = var_fac)) + geom_density()

# boxplot of var_num grouped by var_fact
plot2 <- ggplot(dat %>% filter(estudi=="Normal") %>% filter(complete.cases(t1_n_siv_ec, edat_cat)), aes(x = edat_cat, y = contraste, fill = edat_cat)) + 
  stat_boxplot(geom = "errorbar",
               width = 0.25) + 
  geom_point(aes(fill = edat_cat), shape = 21,
             position = position_jitterdodge()) +
  geom_boxplot(alpha = 0.3)

library(gridExtra)
# Arrange the plots side by side
grid.arrange(plot1, plot2, ncol=2)

# Initialize lblank list
p <- list()

# loop for plot individual graphs and asve in a list
for (i in 1:length(var_long)) {

  datc<- dat_3 %>% filter(Var == var_long[i])

  cat(" \n#####", var_long[i], " \n")

  ### GRAFIC

  p[[i]] <- ggplot(datc, aes_string(x = var_event, y = var_valor)) +

          geom_line(aes_string(group = var_id), alpha = 0.5) +

          stat_summary(aes(group = 1), geom = "line", fun.y = mean,

                       shape = 17, size = 2, col = "red") +

          theme_classic() +

          labs( title = paste("Evolución global", var_long[i]),

                x = NULL,   y = var_valor,    color = NULL)

  print(p[[i]])

  cat("\n")

}
 
# Arrange in a unique plot 
do.call(grid.arrange,p)

#open eoffice library
library(eoffice)

#export ggplot to pptx file, option append is possible
 
topptx(plot2, filename = file.path( "correla.pptx"))

13 Tables

complete_obs <- 90
library(glmnet)

varsurv <- names(dat %>% select(rechazo_agudo:dehiscencia_de_sutura_bronquial,-retiro_de_grapas))

vardel <- NAperc(dat[,varsurv],maxNA = 100 - complete_obs)$var
var_surv <- varsurv[!varsurv %in% vardel]

vart<- names(dat %>% select(tiempo_isquemia_pulmon_derecho,tiempo_isquemia_pulmon_izquierdo,dias_uci,dias_intubacion,retiro_de_grapas,dias_ingreso_hosp,tiempo)) 
# tiempo_ecmo_pre solo tiene 1 valor mayor a 0 

vardel <- NAperc(dat[,vart],maxNA = 100 - complete_obs)$var
var_t <- vart[!vart %in% vardel]

var_surv_ord <- var_surv[order(var_surv)]

sumar <- list()
results <- list()
resul <- matrix()
for (i in seq_along(var_t)) {
  sumar[[i]] <- list()
  results[[i]] <- data.frame()
  rows <- c()
  for (j in seq_along(var_surv_ord))  {
    sumar[[i]][[j]] <- summary.quanti(data = dat,x = var_t[i], group = var_surv_ord[j], show.all = F,show.n = F,var.tidy = F, method = "non-param",
                                 nround = 1)
    #sumar[[i]][[j]]$summary$variable <- " "
    #colnames(sumar[[i]][[j]]$summary)[1]  <- c(label(dat[var_surv_ord[j]]))
    sumar[[i]][[j]] <- t(sumar[[i]][[j]]$summary[,-1])
    sumar[[i]][[j]] <- as.data.frame(sumar[[i]][[j]])
    sumar[[i]][[j]] <- tibble::rownames_to_column(sumar[[i]][[j]], " ")
    rows <- c(rows, rep(label(dat[var_surv_ord[j]]),nrow(sumar[[1]][[j]])))
    results[[i]] <- rbind(results[[i]],sumar[[i]][[j]])
  }
  
  resul <- cbind(resul,results[[i]])
}

r <- resul %>% select(-c(` `,resul))
n <- resul$` `

final_re <- cbind(n,r)
colnames(final_re)[1] <- " "

kk <- final_re %>% kable(longtable = T,escape = F, row.names = F, caption = "Summary of results for time variables",align = "c" )  %>%
  kable_styling(latex_options = c("striped","hold_position", "repeat_header"), full_width = F, fixed_thead = T) %>%
  row_spec(0, color = "white", background = "#993489") %>%
  column_spec(which(names(final_re) == " "), width_max = "9em",bold = T) %>%
  pack_rows(index  = table(rows)) %>%
  add_footnote(escape = F, notation = "symbol" )%>%
  row_spec(which(final_re$` ` == "p.value"), bold = T, align = "right")

print(kk)

14 Printing tricks

Use the knitr chunk option results = "asis" to print the html table cat(tab_model()$knitr)

More elaborated code:

pl <- list()

for (i in seq_along(var_long)) {
  cat(" \n####", var_long[i], " \n")
  
  ### GRAFIC
  print(ggplot(dat3_long, aes_string(x = var_event, y = var_long[i])) +
          geom_line(aes_string(group = var_id), alpha = 0.5) +
          stat_summary(aes(group = 1), geom = "line", fun.y = mean,
                       shape = 17, size = 3) +
          theme_bw() +
          labs( title = paste("Evolución global", var_long[i]),
                x = NULL,   y = var_long[i],    color = NULL))
  
  cat("\n")
  ### MODEL 
  
  ctrl <- lmeControl(opt = 'optim')
  frml <- as.formula(paste0(var_long[i], "~", var_event))
  
  ###############################################################    REVISAR AMB CALMA EFECTES ALEATORIS EN CADA ESTUDI ########################################
  model1 <- lme(frml, random = ~ 1 | id_sujeto, data = dat3_long,
                method = "REML", control = ctrl)
  
  model2 <- lme(frml, random = ~ Tiempo | id_sujeto, data = dat3_long,
                method = "REML", control = ctrl)
  
  if (na.omit(anova(model1, model2)$"p-value") < 0.05) {
    model <- model2
  } else {
    model <- model1
  } 
  
  
  cat(tab_model(model)$knitr,"\n--------\n") # THIS TABLE
  cat("\n")
  print(plot_model(model,colors = "#8d35cc",show.values = T) )
  cat("\n")
}

Simple code:

# need this to use tab_model()
library(sjPlot) 
library(sjmisc)
library(sjlabelled)

data(mtcars)
m.mtcars <- lm(mpg ~ cyl + hp + wt, data = mtcars)
cat(tab_model(m.mtcars)$knitr)