What is Accuracy?
Accuracy refers to the extent to which the observed value matches the true value of the event. In simple terms: Is the number correct? If a child screamed 5 times, but you wrote down 4, your data is inaccurate. For your RBT practice test, remember that Accuracy is the "Gold Standard" of measurement.
In ABA, we strive for accuracy, but human error is common. Accuracy is different from Reliability. You and your supervisor might both agree (Reliability) that the child screamed 4 times, but if the video proof shows 5 times, you are both consistently wrong (Inaccurate).
On the RBT Practice Exam, you might see questions about "Measurement Error." This is the difference between what you recorded and what actually happened.
Sources of Measurement Error
Poorly Designed Measurement Systems: Trying to count high-rate behaviors (like rapid hand flapping) with a frequency count is hard. You will likely miss some, leading to low accuracy.
Observer Expectancy: If you "expect" the child to do better because they started a new medication, you might subconsciously ignore bad behaviors. This is a bias that ruins accuracy.
Clinical Best Practices
To ensure accuracy, we use Calibration. This means comparing your data against a "True Value." Usually, this is done by video modeling. We watch a video where we know the child hits 10 times. If you click your counter 10 times, you are calibrated/accurate.
Continuous Measurement (Frequency, Duration) tends to be more accurate than Discontinuous Measurement (Interval Recording) because it captures every instance.
Accuracy vs. Validity vs. Reliability
The "Big Three" of measurement quality:
- Accuracy: The data is right.
- Validity: The data measures the relevant dimension.
- Reliability: The data is consistent.
RBT Practice Exam Tips for Accuracy
If a question asks: "The BCBA compares the RBT's data to a video recording of the session. What are they checking?" The answer is Accuracy. The video is the "True Value."
Note that "IOA" assesses Reliability, not Accuracy. Two people can agree and still be wrong. Only comparing to a "True Value" verifies accuracy.
Professional Responsibility
Inaccurate data is dangerous. If you record that a child ate 100% of their lunch when they only ate 50%, the parents might think the feeding program is solved. Later, the child loses weight, and it becomes a health crisis.
Honesty is the key to accuracy. If you zoned out and missed a behavior, do not guess. Write "Data Missing" for that interval. A guess is inaccurate data that corrupts the graph.
Always check your tools. Is your stopwatch zeroed out? Is your clicker working? Simple equipment failures are a leading cause of inaccuracies.
Frequently Asked Questions About Accuracy
What is the definition of Accuracy?
The extent to which the observed value matches the true value.
How is Accuracy different from Reliability?
Reliability is about consistency ("Do we agree?"). Accuracy is about truth ("Are we right?").
What is a "True Value"?
A measure obtained by procedures that are independent of the current observation, usually considered the "answer key" (e.g., a video recording counted slowly).
What is Measurement Bias?
Non-random measurement error. Example: An RBT who likes a student might consistently under-report their aggression.
Can data be Reliable but Inaccurate?
Yes. Two observers can both miss the same behavior. They agree (Reliable), but the count is wrong (Inaccurate).
Which measurement is most accurate?
Continuous Measurement (Frequency, Duration) is generally more accurate than Discontinuous Measurement (Interval recording) because it doesn't estimate.
Does IOA prove accuracy?
No. IOA only proves consistency. However, high IOA usually suggests high accuracy, though it doesn't guarantee it.
What is "Calibration"?
The process of correcting your measurement tool or your own observing skills to match a standard "true value."
How do I improve my accuracy?
Minimize distractions. Do not text while observing. Use mechanical tools like counters. Practice with videos.
Why is discontinuous measurement often inaccurate?
Because it is an estimate. Whole Interval recording underestimates behavior. Partial Interval overestimates it. This is a known "measurement artifact."
Is this on the RBT exam?
Yes. You must distinguish between being "correct" (Accuracy) and being "consistent" (Reliability).
What happens if I submit inaccurate data?
It leads to ineffective treatment. The BCBA might keep a program going that isn't working, or stop a program that is needed.