自定义Recode factors转指定数值：NG/SG/LG/MG/HG映射转换失败排查

Hey there! Let's walk through the most common reasons your attempts to recode those Recode factors to custom numeric values are failing, plus the reliable fixes to get it right.

Common Mistakes That Break Your Recoding #

Here are the top pitfalls people hit when trying this kind of factor-to-numeric conversion:

• Directly assigning numeric values to a factor column
  If you tried something like df$Recode <- ifelse(df$Recode == "NG", 0, 1.25) or similar, you’re running into a core quirk of factors: they’re stored as integers with labeled levels. Assigning numeric values directly either forces the column to stay a factor (treating your numbers as new level labels) or mangles the conversion because R uses the underlying integer codes instead of the visible labels.

• Modifying factor levels without converting to numeric
  Changing levels(df$Recode) <- c(0, 1.25, 7.25, 26, 40) makes the labels look right, but the column is still a factor—not a numeric column. If you try to do math on it later, R will use the underlying integer codes (1,2,3,4,5) instead of your custom values.

• Mismatching factor level order
  If your Recode factor’s levels aren’t ordered exactly as NG, SG, LG, MG, HG, assigning levels by position will map the wrong values. For example, if levels are HG, NG, SG, LG, MG, your first level value (0) would get assigned to HG instead of NG.

• Using recode functions incorrectly
  If you used dplyr::recode_factor instead of dplyr::recode, you’re still left with a factor column instead of numeric. Or if you missed a level in your recode calls, you’ll get NA values where you don’t expect them.

Working Solutions to Recode Your Factors #

Try one of these reliable methods to get the numeric values you need:

1. Use dplyr::recode for Exact Mapping #

This method explicitly matches each factor label to your desired numeric value, and returns a numeric column:

library(dplyr)

# Replace df with your actual data frame name
df$Recode_numeric <- recode(df$Recode,
                            NG = 0,
                            SG = 1.25,
                            LG = 7.25,
                            MG = 26,
                            HG = 40)

Make sure you include every factor level in your recode call—any missing levels will result in NA.

2. Create a Mapping Vector #

This approach uses a named vector to map labels to values, which is great if you need to reuse the mapping later:

# Create your mapping (names = factor labels, values = target numbers)
recode_map <- c(NG = 0, SG = 1.25, LG = 7.25, MG = 26, HG = 40)

# Convert factor to character first, then match to the map
df$Recode_numeric <- recode_map[as.character(df$Recode)]

Converting the factor to character is key here—without it, R would use the factor’s underlying integer codes to index the vector, leading to wrong matches if your levels aren’t in order.

3. Use case_when for Flexible Logic #

If you need to add fallback logic (like handling unexpected levels), case_when is a great option:

df$Recode_numeric <- case_when(
  as.character(df$Recode) == "NG" ~ 0,
  as.character(df$Recode) == "SG" ~ 1.25,
  as.character(df$Recode) == "LG" ~ 7.25,
  as.character(df$Recode) == "MG" ~ 26,
  as.character(df$Recode) == "HG" ~ 40,
  TRUE ~ NA_real_ # Assign NA to any unrecognized levels
)

The core takeaway here is always to account for how factors are stored in R—don’t confuse the visible labels with the underlying integer codes. By explicitly matching labels to your target values, you’ll avoid the most common conversion headaches.

内容的提问来源于stack exchange，提问作者HCAI