Summarizing Observation Data: Rate, Mean Duration and Percentage

When you observe a behavior, you end up with raw numbers: tallies, timings, checked boxes. To compare sessions fairly, you often turn those raw numbers into a summary such as a rate, an average length, or a percentage. Every one of these summaries is a division: something you counted or measured, divided by what you counted it against. Most errors come from dividing by the wrong thing.

By the end of this lesson, you should be able to:

• Calculate a rate and explain why a larger count does not always mean a higher rate.
• Calculate mean duration and explain how it differs from total duration.
• Tell apart percentage of time, percentage of opportunities and percentage of intervals.
• Recognize when a calculation is impossible, and why missing data is not zero.

You only need division and multiplying by 100. Each step is shown.

Numerator and denominator

In any fraction, the numerator (top number) is the quantity you counted or measured, such as taps, seconds of crying or correct responses. The denominator (bottom number) is what you counted it against: minutes observed, episodes recorded or opportunities given.

Before calculating, say both parts in words, for example "requests per minute observed." If you can't name the denominator, you aren't ready to divide.

Units matter too. A rate mixes units on purpose: taps divided by minutes gives taps per minute. A percentage is different. It compares a part with a whole of the same kind, so the top and bottom must use the same unit. For example, 30 seconds ÷ 5 minutes = 6 seconds per minute. That is a real quantity, but it is not 6%. Convert first: 5 minutes = 300 seconds, and 30 ÷ 300 × 100 = 10%.

Frequency versus rate

Frequency is how many times the behavior happened. It is a plain count. Rate is the count divided by how long you observed. It tells you how often the behavior happened per minute or per hour.

Rate = count ÷ observation time

Worked example. A learner taps the table 12 times in a 30-minute session on Monday. On Wednesday, there are 18 taps in a 60-minute session.

• Monday: 12 ÷ 30 = 0.4 taps per minute
• Wednesday: 18 ÷ 60 = 0.3 taps per minute

Wednesday has the larger count, but Monday has the higher rate. Wednesday's session was twice as long, so 18 taps were spread over more time. This is why counts from sessions of different lengths cannot be compared directly.

Change one fact. Suppose Wednesday lasted 45 minutes instead. Then 18 ÷ 45 = 0.4 taps per minute, which is the same rate as Monday. The count didn't change. Only the denominator did.

Combining sessions. To get one rate for both original sessions, pool them. Add all the counts, add all the time, then divide once:

(12 + 18) ÷ (30 + 60) = 30 ÷ 90 ≈ 0.33 taps per minute

A tempting shortcut is to average the two rates: (0.4 + 0.3) ÷ 2 = 0.35. That treats the 30-minute session as if it counted as much as the 60-minute session. When sessions have unequal lengths, pool the counts and the time instead.

When there is no rate. If a session is cancelled and no observation happens, the denominator is 0 minutes. You cannot divide by zero, so there is no rate to report. That is different from observing for 30 minutes and seeing no taps. That session has a valid rate of 0 ÷ 30 = 0 taps per minute.

Checkpoint 1

On Tuesday, a learner makes 9 requests in 20 minutes. On Thursday, the learner makes 15 requests in 50 minutes. Which day had the higher rate? What is the combined rate across both days?

Answer: Tuesday; about 0.34 requests per minute combined

Tuesday's rate is 9 ÷ 20 = 0.45 requests per minute. Thursday's rate is 15 ÷ 50 = 0.30 requests per minute. Tuesday is higher even though Thursday has the larger count.

To combine the days, pool the counts and the time: (9 + 15) ÷ (20 + 50) = 24 ÷ 70 ≈ 0.34 requests per minute. Averaging the two rates gives 0.375. That is wrong here because it gives the 20-minute session the same weight as the 50-minute session.

Mean duration

Duration is how long one episode of the behavior lasts. Mean duration is the average length of an episode: add up the episode lengths, then divide by the number of episodes.

Mean duration = sum of episode lengths ÷ number of episodes

Worked example. A learner has three crying episodes lasting 40 seconds, 1 minute 15 seconds and 50 seconds.

1. Convert to one unit. 1 minute 15 seconds = 75 seconds.
2. Add: 40 + 75 + 50 = 165 seconds. This is the total duration.
3. Divide by the number of episodes: 165 ÷ 3 = 55 seconds. This is the mean duration.

The total (165 seconds) answers "How much time did crying take altogether?" The mean (55 seconds) answers "How long was a typical episode?"

A quick check helps here. A mean must fall between the shortest episode and the longest one. 55 seconds sits between 40 and 75, so the answer is reasonable. If you got 165, you reported the total, not the mean.

Notice that the denominator here is the number of episodes, not minutes observed. If no episodes occurred, you have nothing to average. Report that there were no episodes rather than a mean of "0 seconds."

Percentage of observation time

Duration data can also answer a different question: what share of the observation time did the behavior take up?

Percentage of time = time engaged in behavior ÷ total time observed × 100

Using the same three episodes, suppose you observed for 15 minutes. Convert to seconds: 15 × 60 = 900 seconds. Then 165 ÷ 900 × 100 ≈ 18.3% of the observation time.

A common units mistake is to divide 165 seconds by 15 minutes, which gives 11. That 11 means 11 seconds of crying per minute observed. It is not 11%. A percentage needs the top and bottom in the same unit, so convert first. As a check, 11 seconds out of every 60 seconds is 11 ÷ 60 × 100 ≈ 18.3%, which matches.

Percentage of opportunities and percentage of intervals

These are also percentages, but they use different denominators, so they answer different questions.

Percentage of opportunities describes how often a response happened out of the chances to respond. Suppose you give 10 instructions and the learner follows 7 of them. 7 ÷ 10 × 100 = 70% of opportunities. The denominator is the number of instructions given, not the session length.

Percentage of intervals applies when the observation is divided into equal time blocks, called intervals, and each block is scored as yes or no. Suppose 30 intervals are recorded and the behavior is scored in 12 of them. 12 ÷ 30 × 100 = 40% of intervals.

Be careful how you describe this result. "40% of intervals" is not the same as "40% of the time." A scored interval means the behavior met the recording rule for that block. It does not mean the behavior filled the whole block. Also, no percentage tells you why a behavior changed. These numbers describe what was observed, not its cause.

To keep the four summaries straight, ask which question each one answers:

• Rate: How often per unit of time?
• Mean duration: How long was a typical episode?
• Percentage of time: What share of the observed time did the behavior occupy?
• Percentage of opportunities or intervals: Out of the chances or blocks recorded, in how many did it occur?

Checkpoint 2

During a 20-minute observation, a learner has three episodes of leaving the work area: 2 minutes, 90 seconds and 30 seconds. Find the mean duration and the percentage of observation time.

Answer: Mean 80 seconds; 20% of observation time

First convert to seconds: 2 minutes = 120 seconds, and 20 minutes = 1,200 seconds.

• Total duration: 120 + 90 + 30 = 240 seconds.
• Mean duration: 240 ÷ 3 = 80 seconds. Check: 80 falls between 30 and 120.
• Percentage of time: 240 ÷ 1,200 × 100 = 20%.

Dividing 240 seconds by 20 minutes gives 12. That is 12 seconds per minute, not 12%. You need matching units before you can report a percentage. Reporting 240 seconds as the mean confuses the total with the average.

Checkpoint 3

A 10-minute observation is divided into 40 intervals, each 15 seconds long. The behavior is scored in 10 of the intervals. What percentage of intervals is that? Can you report the result as the percentage of time the behavior occurred? Explain.

Answer: 25% of intervals, which is not the same as 25% of the time

The numerator is the number of scored intervals, 10. The denominator is the number of intervals recorded, 40. So 10 ÷ 40 × 100 = 25% of intervals. Dividing by the 10 minutes instead would give 1 scored interval per minute, which is a different quantity and not a percentage.

You should not report this as "25% of the time." A scored interval only tells you the behavior met the recording rule during that 15-second block. In one block it might have lasted 2 seconds; in another it might have filled all 15. Interval data don't record how many seconds the behavior lasted, so they can't give an exact percentage of time. That would require duration data. Report it as "25% of intervals."

Checkpoint 4

You are summarizing a week of data for a skill where the learner is given practice trials:

• Thursday's data sheet is blank because it was misplaced.
• On Friday, 8 trials were presented and the learner responded correctly on none.
• On Saturday, the session ended before any trials were presented.

What should you report for each day?

Answer: Thursday is missing, Friday is 0%, and Saturday has no percentage

• Thursday: The data is missing. Flag it to your supervisor. Do not enter 0%, because nobody knows what happened that day.
• Friday: 0 ÷ 8 × 100 = 0% of opportunities. This is a real, valid zero, because there were 8 chances and no correct responses.
• Saturday: There were zero opportunities, so the denominator is 0. No percentage can be calculated. Record that no trials were presented. Reporting 0% would wrongly suggest the learner failed.

What to remember

• Name the numerator and the denominator in words before dividing.
• A rate deliberately divides by time (such as taps per minute). For a percentage, put the top and bottom in the same unit first.
• Rate = count ÷ time. A bigger count does not always mean a higher rate.
• To combine sessions of unequal length, add up the counts and the time first. Do not average the rates.
• Mean duration = total duration ÷ number of episodes. The mean is not the total.
• Percentage of time, of opportunities and of intervals use different denominators and answer different questions. A percentage of intervals is not a percentage of time.
• A zero denominator means no valid calculation. Missing data must be flagged, not entered as zero.
• Follow your supervisor's recording and rounding conventions. Round only at the final step.

Sources

• Vanderbilt IRIS Center, Measuring Behavior (case study): formulas for rate, percentage of intervals, percentage of time and average duration. https://iris.peabody.vanderbilt.edu/case-study/measuring-behavior/
• Behavior Analyst Certification Board, RBT Test Content Outline (3rd ed.): task A.6, calculating and summarizing data (e.g., rate, mean duration, percentage), and task E.2, seeking clinical direction from a supervisor, including about data irregularities. https://www.bacb.com/wp-content/rbt-outline-3rdEd/
• Behavior Analyst Certification Board, RBT Ethics Code (2.0), standard 2.02 on following supervisor direction and documenting client data accurately. https://www.bacb.com/wp-content/rbt-ethics-code

Independent study resource

This study resource is not affiliated with, sponsored by, or endorsed by the Behavior Analyst Certification Board (BACB). BACB, Registered Behavior Technician and RBT are registered trademarks of the Behavior Analyst Certification Board. Source citations do not imply endorsement by their authors or publishers.