R Academy · Lesson

NA in Calculations and Aggregations

Control NA behavior in mean(), sum(), and other aggregate functions.

NA in Calculations and Aggregations is a free R Academy lesson on CoddyKit — lesson 4 of 4. You can read the complete lesson below for free — then practise it hands-on in the browser with a built-in code editor and a 24/7 AI tutor. It is part of the R Academy learning path, one of 4 lessons in the course, and your progress syncs across the web and the CoddyKit app.

The na.rm Parameter

Most R aggregation functions accept a na.rm argument (NA remove). When set to TRUE, the function ignores NA values before computing the result. By default, na.rm = FALSE, so NAs propagate.

scores <- c(88, NA, 72, 95, NA, 81, 67)
cat('With na.rm=FALSE (default):
')
cat('  mean:', mean(scores), '
')           # NA
cat('  sum: ', sum(scores), '
')            # NA
cat('With na.rm=TRUE:
')
cat('  mean:', mean(scores, na.rm = TRUE), '
')  # 80.6
cat('  sum: ', sum(scores, na.rm = TRUE), '
')   # 403

mean() with na.rm

mean(x, na.rm = TRUE) computes the arithmetic mean of the non-missing values. The denominator is the count of present values, not the total length.

temperatures <- c(22.1, NA, 19.8, 25.1, NA, 17.3, 20.0)

cat('Total elements:', length(temperatures), '
')
cat('Present elements:', sum(!is.na(temperatures)), '
')
cat('Mean (na.rm=TRUE):', mean(temperatures, na.rm = TRUE), '
')

# Manual verification
cat('Manual mean:', sum(temperatures, na.rm=TRUE) / sum(!is.na(temperatures)), '
')

sum() with na.rm

sum(x, na.rm = TRUE) adds only the non-NA values. This is essential for counting TRUE values in a logical vector that may contain NAs, or computing totals from incomplete data.

sales_daily <- c(1200, NA, 980, 1450, NA, 1100, 890)
cat('Sum (default):', sum(sales_daily), '
')            # NA
cat('Sum (na.rm):  ', sum(sales_daily, na.rm = TRUE), '
')  # 5620

# Count observations meeting a condition (NA-safe)
cat('Days over 1000:', sum(sales_daily > 1000, na.rm = TRUE), '
')

min() and max() with na.rm

min() and max() also have the na.rm parameter. Without it, a single NA in the input causes the function to return NA. With na.rm = TRUE, the extremes of the present values are returned.

blood_pressure <- c(120, NA, 135, 118, NA, 142, 125, NA, 130)
cat('Max BP (default):', max(blood_pressure), '
')              # NA
cat('Max BP (na.rm):  ', max(blood_pressure, na.rm = TRUE), '
')  # 142
cat('Min BP (na.rm):  ', min(blood_pressure, na.rm = TRUE), '
')  # 118
cat('Range (na.rm):   ', range(blood_pressure, na.rm = TRUE), '
')

var() and sd() with na.rm

Variance (var()) and standard deviation (sd()) also support na.rm. They compute the statistic using only the non-missing observations, automatically adjusting the degrees of freedom.

exam_scores <- c(78, 85, NA, 92, 68, 75, NA, 88, 91, NA)
n_valid <- sum(!is.na(exam_scores))
cat('Valid observations:', n_valid, '
')
cat('Mean:   ', round(mean(exam_scores, na.rm = TRUE), 2), '
')
cat('SD:     ', round(sd(exam_scores, na.rm = TRUE), 2), '
')
cat('Var:    ', round(var(exam_scores, na.rm = TRUE), 2), '
')

median() and quantile() with na.rm

median() and quantile() are often more robust to outliers than the mean, but they also need na.rm = TRUE to handle missing values correctly.

incomes <- c(45000, NA, 52000, 48000, NA, 210000, 51000, NA, 49000)
cat('Observations:', sum(!is.na(incomes)), '
')
cat('Median income:', median(incomes, na.rm = TRUE), '
')
cat('Mean income:  ', mean(incomes, na.rm = TRUE), '
')  # pulled up by 210000
cat('25th pctile:  ', quantile(incomes, 0.25, na.rm = TRUE), '
')
cat('75th pctile:  ', quantile(incomes, 0.75, na.rm = TRUE), '
')

prod() and cumsum() with na.rm

prod(x, na.rm = TRUE) computes the product of non-NA values. Cumulative functions like cumsum() do NOT have na.rm — NAs propagate from the first missing value onwards.

growth_rates <- c(1.05, NA, 1.03, 1.08, NA, 1.02)
cat('Product of rates:', prod(growth_rates, na.rm = TRUE), '
')

# cumsum propagates NAs
values <- c(10, 20, NA, 40, 50)
cat('cumsum (with NA):', cumsum(values), '
')

# Workaround: replace NA first
filled <- replace(values, is.na(values), 0)
cat('cumsum (NA=0):   ', cumsum(filled), '
')

na.rm in Apply Functions

When using apply() over rows or columns of a matrix, you can pass na.rm = TRUE to the function using .... This computes aggregates while ignoring NAs.

m <- matrix(c(88,NA,72, 95,81,NA, 67,75,90), nrow = 3)
cat('Matrix:
'); print(m)

# Row means ignoring NAs
cat('Row means (na.rm=TRUE):', apply(m, 1, mean, na.rm = TRUE), '
')

# Col means ignoring NAs
cat('Col means (na.rm=TRUE):', apply(m, 2, mean, na.rm = TRUE), '
')

rowSums() / colSums() with na.rm

rowSums() and colSums() have their own na.rm parameter (not passed via ...). Using na.rm = TRUE makes them treat NA as 0 in sums.

sales <- matrix(c(100,NA,80, NA,90,70, 110,85,NA), nrow = 3)
cat('Sales matrix:
'); print(sales)

cat('Row sums (na.rm=F):', rowSums(sales), '
')             # NAs propagate
cat('Row sums (na.rm=T):', rowSums(sales, na.rm = TRUE), '
')
cat('Col means (na.rm=T):', colMeans(sales, na.rm = TRUE), '
')

Comparison: na.rm=FALSE vs TRUE

Understanding when to use na.rm = TRUE vs FALSE is about your analysis intent: FALSE ensures you know data is incomplete; TRUE computes best-estimate statistics from available data.

# Always check: how much data is missing BEFORE using na.rm=TRUE
collected <- c(78, NA, 85, NA, 91, 74, NA, 88, 82, NA)
pct_missing <- mean(is.na(collected)) * 100
cat('Missing:', round(pct_missing, 1), '% of data
')

if (pct_missing < 20) {
  cat('Acceptable: computing mean with na.rm=TRUE
')
  cat('Mean:', round(mean(collected, na.rm = TRUE), 2), '
')
} else {
  cat('Too much missing data — investigate before computing
')
}

na.rm in User Functions

When writing your own aggregation functions, include an na.rm parameter and pass it through to base R functions. This makes your functions consistent with R's conventions.

# Custom function with na.rm support
cv <- function(x, na.rm = FALSE) {
  # Coefficient of Variation = SD / mean
  s <- sd(x, na.rm = na.rm)
  m <- mean(x, na.rm = na.rm)
  return(s / m)
}

readings <- c(22.1, NA, 19.8, 25.1, NA, 17.3, 20.0)
cat('CV (na.rm=FALSE):', cv(readings), '
')          # NA
cat('CV (na.rm=TRUE): ', round(cv(readings, na.rm = TRUE), 4), '
')

Quick Check

What does mean(c(10, 20, NA, 40), na.rm = TRUE) return?

Recap: NA in Calculations

Excellent! Key takeaways from this lesson:

Most aggregation functions (mean, sum, min, max, sd, etc.) have a na.rm parameter
na.rm = FALSE (default) means NA propagates — if any value is NA, the result is NA
na.rm = TRUE ignores NAs and computes the statistic from present values only
The denominator for mean(na.rm=TRUE) is the count of non-NA values, not total length
Always check the proportion of missing data before computing statistics with na.rm=TRUE
Include na.rm in your own aggregation functions to follow R convention

# Complete summary with na.rm
data_vec <- c(55, NA, 72, 88, NA, 61, 79, NA, 93, 68)
cat('N valid:', sum(!is.na(data_vec)), '
')
cat('Mean:  ', round(mean(data_vec, na.rm=TRUE), 2), '
')
cat('Median:', median(data_vec, na.rm=TRUE), '
')
cat('SD:    ', round(sd(data_vec, na.rm=TRUE), 2), '
')
cat('Min:   ', min(data_vec, na.rm=TRUE), '
')
cat('Max:   ', max(data_vec, na.rm=TRUE), '
')

Frequently asked questions

Is the “NA in Calculations and Aggregations” lesson free?

Yes — the full text of “NA in Calculations and Aggregations” is free to read here on the web, and the R Academy course includes 4 lessons in total. To practise it interactively (a built-in code editor and a 24/7 AI tutor) and unlock the rest of the R Academy course, upgrade to CoddyKit PRO.

What will I learn in “NA in Calculations and Aggregations”?

Control NA behavior in mean(), sum(), and other aggregate functions. You practise R Academy with hands-on code you run directly in the browser, and a 24/7 AI tutor answers your questions as you work through the lesson.

Do I need any experience to start R Academy?

No prior experience is required. R Academy on CoddyKit is structured for beginners through advanced learners; this is — lesson 4 of 4, so you can start here or from the beginning and move at your own pace.

How long does the “NA in Calculations and Aggregations” lesson take?

Most CoddyKit lessons take about 5–10 minutes. Each one is bite-sized and interactive, so you make steady progress and pick up exactly where you left off across the web and the app.

Can I write and run code in this R Academy lesson?

Yes. Every R Academy lesson includes a built-in code editor, so you write and run real code right in your browser and get instant AI feedback — no local setup required.

All lessons in this course

← Back to R Academy