Error in y - mu : non-numeric argument to binary operator

Error in y - mu : non-numeric argument to binary operator - linear regression loop

Hi!

I am trying to perform linear regression for age and sex on a dataset with 6 samples and 16757 genes creating a loop. this is my dataset ( I copied the first columns):

'data.frame':   6 obs. of  16757 variables:
 $ samples               : chr  "hu-c_lab13" "hu-c_lab15" "hu-c_lab17" "hu-gent_lab14" ...
 $ treatment             : chr  "untreated" "untreated" "untreated" "treatment" ...
 $ sex                   : chr  "Male" "Female" "Male" "Female" ...
 $ age                   : num  45 56 46 65 21 75
 $ 7SK (i)               : num  87779 79828 64005 44973 42646 ...

I want to do a loop to identify if age and sex affect the gene expression and I wanted to obtain the fitted.values

prova$treatment <- factor(prova$treatment, levels=c("treatment","untreated"))
prova$sex <- factor(prova$sex, levels=c("Female","Male"))
prova$age <- as.numeric(prova$age)

genelist <- prova %>% select(5:16757) #select genes

for (i in 1:length(genelist)) {
  formula <- as.formula(paste("samples ~ ", genelist[i], " + age + sex ", sep=""))
  model <- glm(formula, data = prova)
  print(model[["fitted.values"]])
}

but it gives me

Error in y - mu : non-numeric argument to binary operator

what do I do wrong in the loop?

also if I do for single gene it works:

model2 <- lm(ENSG00000202198 ~ sex + age , data=prova)
summary(model2)
model$fitted.values <- predict(model2)
gene <- model2[["fitted.values"]]
gene  <- as.data.frame(gene)

Thank you

Camilla

bulkrnaseq r linear regression

• 4,429 views • link

updated 5.4 years ago by Sam

• written 5.4 years ago by camillab. •

1 answer

One of your gene might have non-numeric content. You can check that by outputting i in your loo

• 0 views • link

written 5.4 years ago by Sam

like this (apologise I am really bad with loops...)?

   for (gene in 1:length(genelist)) {
  formula <- as.formula(paste("samples ~ ", genelist, " + age + sex ", sep=""))
  model <- glm(formula, data = prova)
  print(model[["fitted.values"]])
}

also I checked if there was any character in the genes am I interested to (sapply(prova[5:16753], class) ) and there is no characters there. may be something to do with the fact that sex is a character?

• 0 views • link

written 5.4 years ago by camillab. •

for (gene in genelist) {
  print(gene)
  formula <- as.formula(paste("samples ~ ", gene, " + age + sex ", sep=""))
  model <- glm(formula, data = prova)
  print(model[["fitted.values"]])
}

Other than non-numeric, you can also check for infinite. This loop should show the name of the last gene that caused the problem, which should allow you to debug the problem. (It is also possible that your samples column is non-numeric)

Sam

• 0 views • link

written 5.4 years ago by Sam