ABA Tau and Baseline-Trend Calculator for Clinicians

An ABA Tau-U baseline trend calculator is useful only when a reviewer can see what was compared, how every pair was signed and which trend component entered the result. This worksheet calculates uncorrected A-versus-B Tau first. It then keeps baseline trend, comparison trend and three declared combination sensitivities in separate rows so the calculator never chooses a correction on the clinician's behalf.

Clinicians & ABA Professionals / Data, Outcomes and Clinical Decision-Making.

What this calculator does

Tau is a rank-based summary built from pairs of observations. For an adjacent A-to-B contrast, each valid baseline value is compared with each valid comparison value. With higher values prespecified as preferred, a B value larger than an A value contributes +1, a smaller B value contributes -1 and a tie contributes 0. The signed total is divided by every cross-phase pair, including ties.

The same sign logic can describe trend within a phase. Each earlier observation is compared with every later observation in that phase. A later value moving in the preferred direction contributes +1, movement in the opposite direction contributes -1 and a tie contributes 0.

Those calculations answer narrow questions about ordering. They do not say how far values moved, whether the outcome matters to the client, whether an intervention caused the change or whether the design demonstrates a functional relation. Keep the graph and the complete visual analysis open beside the sheet.

Use the ABA Tau-U baseline trend calculator as a reproducible sensitivity record, not as a scorecard for choosing the most favorable coefficient.

Lock the analysis before inspecting a preferred result

Start with one outcome in compatible units and one adjacent phase contrast. Record the intended direction before calculating. Preserve chronology even though the uncorrected cross-phase score itself does not use within-phase order.

The locked record should include:

the outcome, operational definition, unit and opportunity or exposure base;
ordered Phase A and Phase B observations with actual dates or sequence positions;
the prespecified preferred direction and why it is meaningful;
phase boundaries, condition labels and the specific contrast under review;
valid, missing, invalid, late, corrected and disputed states;
measurement, intervention, staffing, health, access and setting changes;
the tie rule, sign rule, formula names and worksheet or software version;
the graph version, calculation precision, analyst, reviewer and date;
client and stakeholder priorities, direct feedback and unresolved questions.

Do not enter a missing observation as zero. Do not reverse the preferred direction after seeing the coefficient. If a late or corrected value changes an input, retain the original version and display the revised calculation as a new, traceable analysis.

Copyable ordered input

PhaseOrderDate or sequenceRaw valueUnit and denominatorStatusContext or correction noteA1valid / missing / invalid / late / disputedA2A3A4B1B2B3B4B5

Declare higher values preferred or lower values preferred. If a preferred direction is not clinically or logically defensible, leave the directional score uncalculated.

Uncorrected cross-phase Tau

Let nA be the number of valid baseline observations and nB the number of valid comparison observations. Compare every A value with every B value, for nA x nB total pairs.

For higher values preferred:

+1 when the B value is greater than the A value;
0 when the values tie;
-1 when the B value is less than the A value.

For lower values preferred, reverse the favorable and unfavorable signs. The tie remains zero.

S_AB = favorable AB pairs - unfavorable AB pairs

TauAB = SAB / (nA x nB)

The denominator includes tied pairs even though they contribute zero to the numerator. Require favorable + tied + unfavorable = nA x nB. If either phase has no valid observations, return not calculated, not zero.

Ties contribute zero to signed numerators but remain part of the declared pair denominator.

Within-phase trend scores

For a phase with n valid ordered observations, compare every earlier value with every later value. There are n(n - 1) / 2 time-forward pairs.

S_phase = favorable time-forward pairs - unfavorable time-forward pairs

Tauphase = Sphase / [n(n - 1) / 2]

The sign follows the same prespecified preferred direction as the cross-phase comparison. A positive baseline trend therefore means that Phase A was already moving in the preferred direction. This matters to interpretation, but its presence does not automatically authorize a correction. A phase with fewer than two valid observations has no within-phase trend score.

Keep the score variants separate

The original Tau-U methods paper describes more than one way to combine nonoverlap and trend. The companion method PDF illustrates combinations using signed totals and their corresponding pair-count denominators. This worksheet displays four rows and requires the analyst to name the row used.

Displayed rowSigned numeratorPair denominatorWhat the row asksUncorrected AB TauS_ABD_ABHow do all B values rank against all A values?AB plus B trendSAB + SBDAB + DBWhat is the combined signed ordering across the phase contrast and within B?AB minus A trendSAB - SADAB + DAHow does the display change when the preferred-direction baseline trend is controlled in this declared form?AB plus B trend minus A trendSAB + SB - S_ADAB + DB + D_AWhat is the combined sensitivity when both within-phase components are included?

Here DAB = nA x nB, DA = nA(nA - 1)/2 and D_B = nB(nB - 1)/2. Do not subtract already normalized Tau coefficients when the declared rule calls for combining signed counts. That shortcut uses a different denominator and can produce a misleading value. Do not label every arithmetic variant simply Tau-U; retain the exact row name and formula.

The paper discussed a baseline-trend threshold as an ad hoc convention, not a universal clinical law. This calculator does not apply that threshold, recommend a correction or hide uncorrected Tau. The analyst must state the prespecified method, source and rationale. If an implementation produces a coefficient outside -1 to +1, show review required and inspect signs, denominators and formula selection rather than silently clipping the result.

This page also does not calculate a p-value or confidence interval. If a qualified methods reviewer adds an inferential procedure, preserve its assumptions, variance calculation, tail, software version and planned role separately from the descriptive coefficient.

Blank pair and result ledger

Use one row per cross-phase pair and one row per time-forward pair when verification matters.

ComponentEarlier or A order/valueLater or B order/valueFavorable, tied or unfavorableSignABABA trendB trend

OutputFavorableTiedUnfavorableSigned totalDenominatorCoefficientABA trendB trendAB plus B trendAB minus A trendAB plus B trend minus A trend

The combined rows should preserve their component totals rather than inventing new favorable and unfavorable pair classifications. Reconcile every component denominator independently, retain sufficient precision to reproduce the result and state the final rounding rule.

Imani's fictional sensitivity comparison

Imani is fictional. Her team is reviewing independent responses per matched 15-minute observation with ordinary supports available. Higher values were prespecified as preferred. Phase A contains 3, 4, 4 and 5. Phase B contains 5, 6, 5, 7 and 8. No value represents a real client record.

Across the 4 x 5 = 20 AB pairs, 18 are favorable, 2 are tied and none is unfavorable. Therefore S_AB = 18 - 0 = 18, and uncorrected Tau is 18 / 20 = 0.9000.

Within Phase A there are 4 x 3 / 2 = 6 time-forward pairs. Five are favorable, one is tied and none is unfavorable. Thus S_A = 5 and baseline trend Tau is 5 / 6 = 0.8333.

Within Phase B there are 5 x 4 / 2 = 10 time-forward pairs. Eight are favorable, one is tied and one is unfavorable. Thus S_B = 8 - 1 = 7 and comparison trend Tau is 7 / 10 = 0.7000.

Displayed rowArithmeticResultUncorrected AB Tau18 / 200.9000AB plus B trend(18 + 7) / (20 + 10)0.8333AB minus A trend(18 - 5) / (20 + 6)0.5000AB plus B trend minus A trend(18 + 7 - 5) / (20 + 10 + 6)0.5556

The spread is the point of the sensitivity panel. Phase A already rises in the preferred direction, and different declared combinations answer different mathematical questions. The sheet does not select 0.5000, 0.5556, 0.8333 or 0.9000 as the clinically correct answer. It sends the reviewer back to the planned method, graph, design, measurement and decision question.

What the coefficient leaves out

Rank signs ignore distance. A B value barely above an A value receives the same +1 as a much larger difference. Cross-phase Tau also removes chronology within each phase. A delayed shift, an abrupt boundary change and a gradual drift can share the same AB pair counts.

Missingness, unequal opportunity bases, measurement changes and disputed phase membership can change many pairs at once. Serial dependence and baseline pattern remain visible in the graph even when a single coefficient compresses them. The review of nine nonoverlap techniques supports keeping the specific metric definition attached to the number rather than treating nonoverlap measures as interchangeable.

A 2026 tutorial on visual inference and projection aids emphasizes visual analysis and design features as primary and recommends selecting the focal feature in advance. A straight or rank-based summary can fit variable or nonlinear data poorly. The calculator therefore supplies no universal small, moderate, effective or clinically significant band.

Visual analysis and design remain primary

The What Works Clearinghouse Version 5.0 handbook retains review of level, trend, variability, immediacy, overlap and consistency, together with design and replication requirements. One adjacent phase contrast does not establish experimental control merely because a Tau row is favorable.

Preserve the complete chronological graph, measurement quality, treatment integrity, phase-change logic and repeated demonstrations. Examine alternative explanations, harms, burdens, adverse effects and contextual changes. The SCRIBE 2016 Statement supports clear reporting of single-case methods and results; checklist completeness does not prove study quality or intervention benefit.

Client meaning and accountable professional review

A numeric direction can be preferred in the worksheet while the observed difference is unimportant, unwanted or obtained under unacceptable conditions. Review the outcome, magnitude, burden and tradeoffs with the person through accessible communication. Keep ordinary communication, mobility, sensory and other supports available. Preserve assent, dissent and requested changes when applicable.

The current BACB ethics materials are the professional source for certificants. They do not approve this calculator or turn a coefficient into ethical compliance. Case-specific interpretation and decisions remain with qualified professionals working within competence, licensure, supervision, consent, payer and organizational requirements.

Privacy, security and version history

Use synthetic or de-identified values for training whenever possible. When identifiable information is legitimately required, use the least identifying data needed, store the worksheet only in approved systems, limit access and retain correction and export history. HHS provides a current Privacy Rule summary and Security Rule summary for regulated entities. Neither page certifies a particular worksheet or determines whether HIPAA applies to a specific workflow.

Close the record with the selected formula, uncorrected and sensitivity results, graph version, disagreements, client input, qualified interpretation, decision owner, unresolved evidence and next review date. Reopen it when a value, phase boundary, preferred direction, formula, condition or question changes.

ABA Tau and Baseline-Trend Correction Sensitivity Calculator for Clinicians