Compare independently recorded counts, trials, intervals or durations using named formulas, aligned windows and visible disagreements.
How to use this tool
Use this observer agreement calculator to practice comparing independently recorded paired observations using the appropriate agreement method.
Select the method that matches how both observers collected the data.
Enter the two observers’ records for the same observation windows, trials or intervals.
Review valid pairs, excluded or missing pairs and the denominator before exporting.
Fictional example and how to read the result
Fictional total-count example: observer A records 8 events and observer B records 10. Total-count agreement is 80%; this does not prove event-by-event agreement.
Agreement describes correspondence between the entered records. It is not accuracy, reliability certification or evidence that a goal is clinically valid.
Choose the formula that answers your recording question
Observers must collect the same events, trials or windows independently. Identifiers and units must match exactly before a pair is included. Compare before reconciling differences.
Count methods use whole event counts. Duration methods use nonnegative amounts in the same declared time unit. Trial and interval methods use binary 0/1 values, with explicit shared scoring rules.
Total-count and total-duration 0/0 are undefined. Mean count/duration-per-pair treats a paired 0/0 as 100% agreement. Exact/trial/interval methods count paired zeroes as exact matches. Scored/unscored interval methods require at least one qualifying pair.
Each window/unit combination has its own result. Distinct windows or units are never collapsed into one unexplained score. High agreement does not prove accuracy, and matching totals can hide disagreements in individual intervals.
Total count/duration: smaller summed value ÷ larger summed value.
Mean count/duration per pair: average each smaller/larger ratio, including stated zero-zero handling.
Scored interval: both 1 ÷ either 1. Unscored interval: both 0 ÷ either 0.
No universal pass/fail threshold is applied. Missing pairs are counted and excluded.
What to enter in each field
Agreement method: Choose the formula for your records: count uses whole events, duration uses time, and trial/interval uses 0 or 1. Total, mean and exact agreement answer different questions; no pass/fail threshold is applied.
Shared observable definition: Write the shared observable event or scoring rule used by both observers. Agreement on different definitions is not a valid comparison.
Independent recording and alignment confirmation: Describe how both observers independently recorded the same numbered events/windows before comparing records. The tool checks declared matching labels, not whether independence actually occurred.
Paired trial / interval ID: Use a distinct paired event, trial or interval ID within each observation window. Put both observers’ values for that same opportunity on one row.
Observer A window ID: Enter the anonymous window label used by observer A, such as window 1. It must match observer B’s window exactly for inclusion.
Observer B window ID: Enter the same window label when both observed the same period. A mismatched window is disclosed and excluded from the agreement result.
Observer A unit: State observer A’s unit, such as events, seconds or 0/1 intervals. Both observers must use exactly the same unit label.
Observer B unit: State observer B’s unit; it must match observer A. Convert durations before entry rather than mixing seconds with minutes.
Observer A recorded value: Enter observer A’s independently recorded value: whole count, nonnegative duration or binary 0/1 as required by the chosen method. Blank is missing; 0 is an observed zero.
Observer B recorded value: Enter observer B’s value for the same pair using the same unit. Leave blank if unrecorded. Zero/zero handling depends on the named method and is explained in the results.
Original fictional example and data fields
Use the editable blank CSV or local import to collect your own de-identified entries. The example below is fictional.
Observer-agreement calculator
Paired trial / interval ID
Observer A window ID
Observer B window ID
Observer A unit
Observer B unit
Observer A recorded value
Observer B recorded value
1
Fictional window A
Fictional window A
events
events
2
3
2
Fictional window A
Fictional window A
events
events
0
0
3
Fictional window A
Fictional window A
events
events
4
4
Professional sources and permissions
Original worksheets and arithmetic support recording and planning. Follow the official publisher routes for standardized instruments.
Your browser, your entries
Entries are processed in this browser. These pages do not run marketing analytics or send entries to TherapyCRM. Download your work before leaving. Any optional device storage must be selected explicitly.