tidyr drop_na() in R: Remove Rows With NA Values

The drop_na() function in tidyr removes rows with any NA values. Specify columns to limit which NAs trigger row removal. It is the pipeline-friendly alternative to base R na.omit() and complete.cases().

drop_na(df)                              # remove rows with ANY NA
drop_na(df, x)                           # remove rows where x is NA
drop_na(df, x, y)                        # remove rows where x OR y is NA
drop_na(df, where(is.numeric))           # NA in any numeric column
drop_na(df, starts_with("score"))        # NA in score-prefixed cols
filter(df, complete.cases(df))           # equivalent base R
filter(df, !is.na(x))                    # one column, base-style condition

What drop_na() does in one sentence

drop_na() returns a data frame with rows removed wherever specified columns contain NA. Without arguments, it drops rows with NA in ANY column; with column names, it drops rows where those specific columns have NA.

This is one of the most-used data cleaning steps. Use it whenever an analysis cannot tolerate missing values and you have decided that complete-case analysis is acceptable.

Syntax

drop_na(data, ...) is the form. The ... argument selects columns to check for NA.

Five common patterns

1. Drop rows with any NA

The most aggressive form. Equivalent to base R na.omit(df) or df[complete.cases(df), ].

2. Drop rows where one column is NA

NAs in column y are KEPT; only rows with NA in x are dropped. Useful when only specific columns are required for the next step.

3. Drop based on multiple columns

drop_na(x, y) removes rows where x OR y is NA.

4. Drop based on column predicate

Tidyselect helpers like where(), starts_with(), matches() work inside drop_na for flexible column selection.

5. Drop in a pipeline

drop_na() integrates cleanly into dplyr pipelines, unlike na.omit() which can break flow.

drop_na() vs na.omit() vs complete.cases()

When to use which:

• Use drop_na() in dplyr/tidyverse pipelines.
• Use na.omit() for one-line scripts on the whole data frame.
• Use complete.cases(df[, c("x", "y")]) inside filter() for very specific column subsets.

Common pitfalls

Pitfall 1: dropping more rows than expected. drop_na() without arguments removes rows with NA in ANY column. With many columns, this can drop most of your data. Always check nrow(df) before and after.

Pitfall 2: hiding the missingness pattern. Silently dropping rows with NA can make patterns invisible. Before dropping, summarize: summarise(across(everything(), ~ sum(is.na(.)))) shows how many NAs per column.

Try it yourself

Related tidyr / dplyr functions

After mastering drop_na, look at:

• replace_na(): replace NA with a specific value
• fill(): forward/backward-fill missing values
• coalesce(): take first non-NA across columns
• complete(): fill in implicit missing combinations
• naniar::miss_var_summary(): NA summary by column
• mice::mice(): multiple imputation for missing data

For "carry forward last observation" patterns common in time series, tidyr::fill() is the right tool, not drop_na().

FAQ

How do I remove rows with NA in R?

Use tidyr::drop_na(df) to remove rows with NA in ANY column. Use drop_na(df, x) to remove rows where only column x is NA. Base R alternatives: na.omit(df) (whole data frame) or df[complete.cases(df), ].

What is the difference between drop_na and na.omit?

drop_na() is from tidyr and works inside pipelines. It accepts column arguments to limit which NAs trigger row removal. na.omit() is base R, always drops rows with NA in any column, and is harder to chain in pipelines.

How do I drop NA rows for specific columns only?

Pass column names to drop_na: drop_na(df, x, y) removes rows where x OR y is NA. NAs in other columns are kept.

Should I drop NA rows or impute them?

Drop when missingness is small and completely random. Impute when missingness is systematic and you cannot afford to lose rows. For complex missingness, multiple imputation via mice package is the gold standard.

Can I use drop_na with tidyselect helpers?

Yes. drop_na(df, where(is.numeric)) drops rows where any numeric column is NA. drop_na(df, starts_with("score")) checks score-prefixed columns. All standard tidyselect helpers work.