The Foundation of Data-Based Decision Making
Visual analysis is the primary method behavior analysts use to evaluate the effects of interventions on behavior. Unlike statistical analysis used in group research designs, single-subject research in ABA relies on visual inspection of graphed data to determine whether a functional relationship exists between the independent variable and the dependent variable. Mastering visual analysis is essential for both clinical practice and the BCBA exam.
Visual analysis involves examining data within a condition (within-phase analysis) and between adjacent conditions (between-phase analysis). Three core dimensions guide this process: level, trend, and variability. Understanding these dimensions and how they interact allows you to draw accurate conclusions about the effectiveness of your interventions.
Level: The Central Tendency of a Data Series
Level refers to the value around which a series of data points converges. It is essentially the average or typical performance within a given phase. When you look at a graph, level tells you "how much" of the behavior is occurring during that condition.
Calculating Level
You can estimate level by calculating the mean or median of the data points within a phase. The mean is the arithmetic average and is appropriate when data are relatively stable. The median is the middle value and is preferred when there are outliers that could distort the mean. Many practitioners draw a horizontal line through the data at the level value to make comparisons between phases easier to visualize.
Evaluating Level Change Between Phases
When comparing level between adjacent phases, look for a clear shift in the data. If you implement an intervention and the level of the target behavior changes in the expected direction, this provides initial evidence that the intervention may be effective. The magnitude of the level change matters: a large, abrupt shift in level between baseline and intervention provides stronger evidence than a small, gradual change.
Trend: The Direction and Slope of Data
Trend describes the direction and rate of change in the data over time within a phase. A trend can be increasing (accelerating), decreasing (decelerating), or zero (flat). The slope of the trend line indicates how rapidly the behavior is changing.
Methods for Estimating Trend
The split-middle method is a systematic approach to drawing a trend line through a data series. To use this method, divide the data points in a phase into two equal halves. Find the median data value and the median session number for each half. Plot these two intersection points and draw a line connecting them. This line represents the trend of the data.
You can also estimate trend through freehand inspection by drawing a line of best fit through the data. While less precise, this method is often used in clinical practice when quick visual decisions are needed.
Trend and Interpretation
Trend within the baseline phase is particularly important for interpreting intervention effects. If the baseline shows a therapeutic trend (the data are already improving before the intervention), it becomes difficult to attribute subsequent improvement to the intervention. Ideally, baseline data should be stable or contra-therapeutic (moving in the undesired direction) before introducing an intervention, so that any change in trend following intervention introduction can be more confidently attributed to the treatment.
Variability: The Spread and Consistency of Data
Variability refers to the degree to which individual data points deviate from the overall level and trend of the data series. Low variability means data points cluster tightly around the level or trend line. High variability means data points are widely scattered.
Why Variability Matters
Variability directly impacts your ability to draw conclusions from visual analysis. When data are highly variable, it is harder to detect meaningful changes between conditions because the "noise" in the data may obscure the "signal" of the intervention effect. Stable data with low variability provide a clearer basis for comparison and stronger evidence for or against intervention effectiveness.
Addressing Variability
If baseline data show high variability, consider whether measurement procedures are reliable, whether there are uncontrolled variables influencing the behavior, or whether the session length or observation conditions are inconsistent. Sometimes extending the baseline phase allows variability to stabilize. Improving interobserver agreement (IOA) can also reduce measurement-related variability.
Between-Phase Comparison: Putting It All Together
Effective visual analysis requires evaluating all three dimensions simultaneously when comparing adjacent phases. Here is the process you should follow when analyzing data between conditions:
- Compare levels - Did the average level of behavior shift between phases? How large was the shift?
- Compare trends - Did the direction or slope of the data change when the intervention was introduced? A change in trend direction provides stronger evidence than a change in level alone.
- Evaluate variability - Are the data stable enough within each phase to support a meaningful comparison? Overlapping data points between phases weaken the demonstration of an effect.
- Assess immediacy of change - Did the change occur immediately upon introduction of the new condition, or was there a delayed effect? Immediate changes provide stronger evidence of a functional relationship.
- Check for overlap - Do data points in the intervention phase fall within the range of data points in the baseline phase? Less overlap indicates a stronger intervention effect.
Common Patterns on the BCBA Exam
The BCBA exam frequently presents graphs and asks you to interpret them. You may be asked to identify whether a functional relationship is demonstrated, select the best description of the data, or determine what action to take based on the data pattern. Practice with these common scenarios:
Clear Demonstration of Effect
Stable baseline with low variability followed by an immediate, large change in level and trend upon intervention introduction with minimal overlap. This pattern strongly supports the conclusion that the intervention was effective.
Ambiguous Results
Baseline with a therapeutic trend, small level change upon intervention, high variability, and substantial overlap between phases. This pattern does not clearly demonstrate an intervention effect and warrants continued data collection or modification of the intervention.
Delayed Effect
No immediate change upon intervention introduction, but a gradual shift in level or trend over several sessions. This pattern may indicate that the intervention takes time to produce effects, but it weakens the demonstration of experimental control because other variables may be responsible for the delayed change.
Tips for Exam Preparation
Practice visual analysis by reviewing published single-subject research articles and examining the graphs before reading the results section. Try to identify level, trend, and variability in each phase and determine whether you would conclude that a functional relationship was demonstrated. Compare your analysis with the authors' conclusions. The more graphs you analyze, the more confident you will become in your ability to interpret data on the exam.