How to compute a new variable in R based on values of multiple other variables?

Question

I am trying to compute a new variable based on values of multiple existing variables with conditional computation. Specifically, the new variable is kidney function (eGFR), which is estimated based on one's sex, age, being (non-)Negroid, and the concentrations of two blood compounts (i.e., creatinine and cystatin C).

I have tried to accomplish this with R's if...else statement, but encountered a warning message, after which nothting happens. All variables are contained within data frame 'd'.

Basically, what I would like R to do is: if a subject is of the male sex (== 1) and non-Negroid (!= 1), has a blood creatinine ≤ 0.9 and cystatin C ≤ 0.8, then one's kidney function is estimated through:

https://latex.codecogs.com/png.latex?\bg_white&space;eGFR=135\cdot\left&space;(&space;\frac{creatinine}{0.9}&space;\right&space;)^{-0.207}\cdot\left&space;(&space;\frac{cystatinC}{0.8}&space;\right&space;)^{-0.375}\cdot0.995^{age}

and so forth. To this purpose I applied the following piece of code:

if (d$sex == 1 & d$creatinine <= 0.9 & d$cystatinC <= 0.8 & d$race != 1){ ### Non-Negroid males
    d$eGFR <- 135 * I((d$creatinine / 0.9)^-0.207) * I((d$cystatinC / 0.8)^-0.375) * I((0.995)^d$age)
  } else if (d$sex == 1 & d$creatinine <= 0.9 & d$cystatinC > 0.8 & d$race != 1){
    d$eGFR <- 135 * I((d$creatinine / 0.9)^-0.207) * I((d$cystatinC / 0.8)^-0.711) * I((0.995)^d$age)
  } else if (d$sex == 1 & d$creatinine > 0.9 & d$cystatinC <= 0.8 & d$race != 1){
    d$eGFR <- 135 * I((d$creatinine / 0.9)^-0.601) * I((d$cystatinC / 0.8)^-0.375) * I((0.995)^d$age)
  } else if (d$sex == 1 & d$creatinine > 0.9 & d$cystatinC > 0.8 & d$race != 1){
    d$eGFR <- 135 * I((d$creatinine / 0.9)^-0.601) * I((d$cystatinC / 0.8)^-0.711) * I((0.995)^d$age)
  } else if (d$sex == 0 & d$creatinine <= 0.7 & d$cystatinC <= 0.8 & d$race != 1){ ### Non-Negroid females
    d$eGFR <- 130 * I((d$creatinine / 0.7)^-0.248) * I((d$cystatinC / 0.8)^-0.375) * I((0.995)^d$age)
  } else if (d$sex == 0 & d$creatinine <= 0.7 & d$cystatinC > 0.8 & d$race != 1){
    d$eGFR <- 130 * I((d$creatinine / 0.7)^-0.248) * I((d$cystatinC / 0.8)^-0.711) * I((0.995)^d$age)
  } else if (d$sex == 0 & d$creatinine > 0.7 & d$cystatinC <= 0.8 & d$race != 1){
    d$eGFR <- 130 * I((d$creatinine / 0.7)^-0.601) * I((d$cystatinC / 0.8)^-0.375) * I((0.995)^d$age)
  } else if (d$sex == 0 & d$creatinine > 0.7 & d$cystatinC > 0.8 & d$race != 1){
    d$eGFR <- 130 * I((d$creatinine / 0.7)^-0.601) * I((d$cystatinC / 0.8)^-0.711) * I((0.995)^d$age)
  } else if (d$sex == 1 & d$creatinine <= 0.9 & d$cystatinC <= 0.8 & d$race == 1){ ### Negroid males
    d$eGFR <- 145.8 * I((d$creatinine / 0.9)^-0.207) * I((d$cystatinC / 0.8)^-0.375) * I((0.995)^d$age)
  } else if (d$sex == 1 & d$creatinine <= 0.9 & d$cystatinC > 0.8 & d$race == 1){
    d$eGFR <- 145.8 * I((d$creatinine / 0.9)^-0.207) * I((d$cystatinC / 0.8)^-0.711) * I((0.995)^d$age)
  } else if (d$sex == 1 & d$creatinine > 0.9 & d$cystatinC <= 0.8 & d$race == 1){
    d$eGFR <- 145.8 * I((d$creatinine / 0.9)^-0.601) * I((d$cystatinC / 0.8)^-0.375) * I((0.995)^d$age)
  } else if (d$sex == 1 & d$creatinine > 0.9 & d$cystatinC > 0.8 & d$race == 1){
    d$eGFR <- 145.8 * I((d$creatinine / 0.9)^-0.601) * I((d$cystatinC / 0.8)^-0.711) * I((0.995)^d$age)
  } else if (d$sex == 0 & d$creatinine <= 0.7 & d$cystatinC <= 0.8 & d$race == 1){ ### Negroid females
    d$eGFR <- 140.4 * I((d$creatinine / 0.7)^-0.248) * I((d$cystatinC / 0.8)^-0.375) * I((0.995)^d$age)
  } else if (d$sex == 0 & d$creatinine <= 0.7 & d$cystatinC > 0.8 & d$race == 1){
    d$eGFR <- 140.4 * I((d$creatinine / 0.7)^-0.248) * I((d$cystatinC / 0.8)^-0.711) * I((0.995)^d$age)
  } else if (d$sex == 0 & d$creatinine > 0.7 & d$cystatinC <= 0.8 & d$race == 1){
    d$eGFR <- 140.4 * I((d$creatinine / 0.7)^-0.601) * I((d$cystatinC / 0.8)^-0.375) * I((0.995)^d$age)
  } else if (d$sex == 0 & d$creatinine > 0.7 & d$cystatinC > 0.8 & d$race == 1){
    d$eGFR <- 140.4 * I((d$creatinine / 0.7)^-0.601) * I((d$cystatinC / 0.8)^-0.711) * I((0.995)^d$age)
  }

However, when running this R yields:

Warning message:
In if (d$sex == 1 & d$creatinine <= 0.9 & d$cystatinC <= 0.8 &  :
  the condition has length > 1 and only the first element will be used

Anyone who can help me out?

UPDATE: Below are some sample data, including age, sex (0=female, 1=male), race (1=Negroid, != 1 being non-Negroid), creatinine, cystatin C, and the manually calculated eGFR for formula verification purposes:

reconstruct <- structure(list(sex = structure(c(2L, 2L, 2L, 2L, 1L, 1L, 2L, 
2L, 1L, 2L), .Label = c("0", "1"), class = "factor"), race = structure(c(1L, 
1L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 3L), .Label = c("0", "1", "2", 
"3", "4"), class = "factor"), age = c(71.9425051334702, 65.1964407939767, 
46.2258726899384, 51.7152635181383, 54.8747433264887, 71.6714579055441, 
36.0793976728268, 54.3764544832307, 57.9110198494182, 49.9438740588638
), creatinine = c(0.633484162895928, 0.984162895927602, 0.769230769230769, 
0.8710407239819, 0.769230769230769, 0.690045248868778, 0.893665158371041, 
1.02941176470588, 0.83710407239819, 0.701357466063348), cystatinC = c(0.73, 
0.85, 0.64, 0.9, 0.83, 0.95, 1.04, 1, 0.95, 0.68), eGFR =     c(96.1605293085191, 
73.17567750685, 105.934761135043, 80.8974371814808, 103.186483803272, 
88.1306212690947, 77.7383905116244, 66.9892381719287, 90.7223944432609, 
107.443909414004)), row.names = c(NA, 10L), class = "data.frame")

you are comparing a whole vector with one value. this returns a vector like [false true true false]. At the end you ask if([false true true false]), which does not make sense. So R cuts this to the first value. You need to do other logical operators. Try %in% or something like this, which returns just one logical value. — mischva11
– mischva11, Commented Jul 6, 2019 at 12:21
you could also try use any() around all your conditions, but this fixes your syntax. try to think about vectorized operations like ifelse — mischva11
– mischva11, Commented Jul 6, 2019 at 12:29
You could use the ifelse() function rather than the if ... else construct -- but with so many conditions that would be neither efficient nor readable. — John Coleman
– John Coleman, Commented Jul 6, 2019 at 12:29
Perhaps you could first create a new factor variable based on d$sex and d$race and develop simplified formulas for the resulting 4 factor levels (formulas that can be abstracted to their own function definitions) and then populate the new variable in a more readable way by using this factor and the new functions. — John Coleman
– John Coleman, Commented Jul 6, 2019 at 12:45
Can you post sample data, preferably covering all cases in dput format? If yes,dit the question with the output of dput(head(d, 20)). — Rui Barradas
– Rui Barradas, Commented Jul 6, 2019 at 12:56

Rui Barradas · Accepted Answer · 2019-07-08 10:42:58Z

2

I believe the function below follows what is defined in the question but it is untested, since there are no data and expected output.

eGFRfun <- function(DF){
  i_sex <- DF[["sex"]] == 1
  i_creat_0.9 <- DF[["creatinine"]] <= 0.9
  i_creat_0.7 <- DF[["creatinine"]] <= 0.7
  i_cyst <- DF[["cystatinC"]] <= 0.8
  i_race <- DF[["race"]] == 1

  const_fac <- ifelse(i_race, 135, 145.8) + 5*(i_sex - 1)
  creat_denom <- ifelse(i_sex, 0.9, 0.7)
  creat_pow <- ifelse(i_sex & i_creat_0.9, -0.207, -0.601)
  creat_pow <- ifelse(i_sex & i_creat_0.7, -0.248, -0.601)
  cystC_fac <- (DF[["cystatinC"]] / 0.8)^ifelse(i_cyst, -0.375, -0.711)
  age_fac <- 0.995^DF[["age"]]

  const_fac * (DF[["creatinine"]] / creat_denom)^creat_pow * cystC_fac * age_fac
}

Example usage:

d$eGFR <- eGFRfun(d)

edited Jul 8, 2019 at 10:42

answered Jul 6, 2019 at 14:01

Rui Barradas

78k8 gold badges41 silver badges75 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

How to compute a new variable in R based on values of multiple other variables?

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related