How to read data with correct types in R?

Wrong data type can be a serious headache sometimes. It is something easy to overlook until some bugs jump into your sight. Here I will summarise several ways to avoid this type of mistakes.

1. numeric type changes to character. Sometimes, there are weird data mingled in your numeric data, such as “-“, or “1 not”. Sometimes, it can be obvious, sometimes not. as.numeric solves most of them. It may cause warnings, so you may want to wrap with suppresswarnings(as.numeric()).

   You can avoid it beforehand when reading data in. In read.csv, you can use colClasses argument to specify. It actually can reduce your loading time. The possible values are “NA”(default), “NULL”(skipped), “logical”, “integer”, “numeric”, “complex”, “character”, “raw”, “factor”, “Date”, or “POSIXct”. For missing value, na.strings argument can be specified. It can be a vector of strings. Blank fields are automatically considered to be missing no matter it is specified or not.

   In read_excel, col_types can be specified. “NULL” means guessing all columns. A vector is used to speficy each column. The possible values can be “skip”, “guess”, “logical”, “numeric”, “data”, “text” or “list”. For missing values, na argument can be specified. By default, it is “”.

2. character changes to factor type. This sometimes can cause serious headache. Pre-specify the type when readding data is a good habit. In read.csv, the characters are converted to factor by default. If conversion is not desired, set stringAsFactors as FALSE, or set colClasses specifically.

   In read_excel, characters will stay as characters.

   One difference I noticed of factor vs character is when converting to vector. For example, char1 is a character column. char2 = factor(char1). Then c(char1) will give a list of characters, whereas c(char2) will give a list of numbers.

   unique(char1) will give the unique strings in char1, levels(char2) will give the same thing. c(unique(char1)) is equal to c(levels(char2)), but distinct to c(unique(char2)).