R: dplyr conditionals for filter() inside a loop making ANOVA

Question

I'm struggling to create a function that makes ANOVA analysis for a set of dataframes returning different number of objects to the global environment depending on wheter I want to split between the positive/negative trend values (4 objects expected) or not (2 objects expected).

My data is like this:

set.seed(88)
A <- data.frame(ID = 1:100, na.cnt.2 = round(rnorm(n=100,mean = 24.50, sd=5.877722),0),
       trend = rnorm(100, mean = -0.0029446, sd=0.004951971),
       p.val= rnorm(100, mean = 0.1983439, sd=0.2747593))
B <- data.frame(ID = 1:100, na.cnt.2 = round(rnorm(n=100,mean = 22.40, sd=6.180722),0 ),
             trend = rnorm(100, mean = -0.0030070, sd=0.005016312),
             p.val= rnorm(100, mean = 0.1657485, sd=0.3297854))

This is my main code:

anova_subgroups <- function(mp= 0.5, p_val = 0.05, split.sample=FALSE){

n <- 0
for (df in list(A, B)){
 for (name in c("pos", "neg")){
  
index <- ifelse(n==0,"A","B")
letter <- "Myletter"


inner_sample.n <- df %>%  mutate(mpe = na.cnt.2/49) %>% 
  {if (split.sample==TRUE) {if(name =="pos") dplyr::filter(mpe<mp, p.val<p_val, trend>0)
    else dplyr::filter(mpe<mp, p.val<p_val, trend<0)}
    else dplyr::filter(mpe<mp, p.val<p_val)} %>%
  nrow()

alpha = 0.05

degrees.freedom = inner_sample.n - 1

t.score = qt(p=alpha/2, df=degrees.freedom, lower.tail = F)

inner_mean <- df %>%  mutate(mpe = na.cnt.2/49) %>%
  {if (split.sample==TRUE) {if(name =="pos") 
    dplyr::filter(mpe<mp, p.val<p_val, trend>0)
    else dplyr::filter(mpe<mp, p.val<p_val, trend<0)}
    else dplyr::filter(mpe<mp, p.val<p_val)}  %>%
  summarise(mean(trend)) %>%  as.numeric()


inner_sd <- df %>%  mutate(mpe = na.cnt.2/49)%>%
  {if (split.sample==TRUE) {if(name =="pos") 
    dplyr::filter(mpe<mp, p.val<p_val, trend>0)
    else dplyr::filter(mpe<mp, p.val<p_val, trend<0)}
    else dplyr::filter(mpe<mp, p.val<p_val)} %>%
  summarise(sd(trend)/sqrt(inner_sample.n)) %>%  as.numeric()

margin.error <- t.score * inner_sd


lower.bound <- inner_mean - margin.error
upper.bound <- inner_mean + margin.error

CI <- c(lower.bound, upper.bound)
data <- df %>% mutate(mpe = na.cnt.2/49) %>%
  {if (split.sample==TRUE) {if(name =="pos") 
  dplyr::filter(mpe<mp, p.val<p_val, trend>0)
  else dplyr::filter(mpe<mp, p.val<p_val, trend<0)}
  else dplyr::filter(mpe<mp, p.val<p_val)} 

if (split.sample== FALSE){
  
  assign(paste0(index,"_", letter, "_CI"), CI, envir = globalenv())
  assign(paste0(index,"_", letter, "_MEAN"), inner_mean, envir = globalenv())
  assign(paste0(index,"_", letter, "_DF"), data, envir = globalenv())
  break
} else {
  assign(paste0(name,"_",index,"_",letter, "_CI"), CI,
         envir = globalenv())
  assign(paste0(name,"_", index, "_", letter, "_MEAN"), inner_mean, 
         envir = globalenv())
  assign(paste0(name,"_", index, "_", letter, "_DF"), data,
         envir = globalenv())
}

}
n <- n+1
}
}

I'm getting the following error when using the anova_subgroups() function:

Error in dplyr::filter(mpe < mp, p.val < p_val) : object 'mpe' not found

I tried to solve the problem using this approach. That's why in the code you can find the dplyr:: filter() form. Also I'm using this approach to make the conditionals for the filter section.

I even tried to debug the problem by myself making a small version of the logic in the code:

  a <- matrix(-10:10, ncol = 1) %>%  as.data.frame()
  names(a) <-  "v"
  split <- TRUE

 for(df in c("A", "B")){
 for (name in c("pos","neg")){
 a %>% mutate(new= v/49) %>% {if (split==TRUE) {if(name=="pos") 
                        filter(., new>0) 
                      else filter(., new<0)} 
    else filter(., new==10)}
 if (split == FALSE){
 print(paste0(df, a))
 break 
 } else{  print(paste0(df, name, a))
 } 

Getting the following output:

[1] "Apos-10:10"
[1] "Aneg-10:10"
[1] "Bpos-10:10"
[1] "Bneg-10:10"

I thought that my problem was related to the combination of both dplyr::filter() and the {if (something) else} statements. However it seems not to be the case. I really can't understand where the problem relies.

Can anyone explain and guide me to a solution?

Answer 1

I found my problem. I was needing to to add a . inside the filter() statement as following:

data <- df %>% mutate(mpe = na.cnt.2/49) %>%
{if (split.sample==TRUE) {if(name =="pos") 
dplyr::filter(., mpe<mp, p.val<p_val, trend>0)
else dplyr::filter(., mpe<mp, p.val<p_val, trend<0)}
else dplyr::filter(., mpe<mp, p.val<p_val)} 

I think that adding this point is to specify to the filter function that look for the variable inside the main data being used (df in this case).