You need to collect enough data for a stable pattern to be demonstrated. To obtain a reliable baseline, it is ideal to collect a minimum of 3-5 baseline data points. The more data points collected, the more reliably we can predict what would occur if we do not intervene. In practice, it is not always feasible to collect extended baseline data; incorporating data sources and methods that have documented reliability predict behavior (i.e., DIBELS, etc.) can increase consistency in baseline with fewer baseline data points.

Baseline data at its core is collected to provide prediction about a behavior; if nothing is changed, what would the behavior pattern look like. If a stable baseline pattern can be collected across three days (i.e., three days of zero desire behavior or three days of 6 instances of aggression), intervention can be initiated. However, it is important to establish a pattern of behavior to make a defensible argument regarding intervention effectiveness. It is therefore necessary to take sufficient time to collect baseline data that provides a clear pattern in behavior to use as a comparison point once intervention is initiated.

While it may be tempting to begin intervention immediately, it is necessary to collect a series of data points in baseline to make an argument about intervention effectiveness. If a stable behavior is being measured, one could argue that three baseline data points is a defensible amount to document consistent and reliable data. After three data points are collected, you may begin intervention.

The amount of time to pass between data collection in a single subject design largely depends on the behavior and context in which it is being observed. For instance, targeting frequent and severe behavior, it may be necessary to collect multiple data points in a one-hour setting because we would expect with time intensive intervention, the behavior would respond to intervention quickly and change in topography. We would not expect a student receiving a small group reading intervention to increase fluency across a ten-minute period, therefore twice weekly progress monitoring or data collection may be more appropriate. Goal lines can guide the predicted duration of the intervention and therefore the intervals by which to collect data.

In order to determine if your intervention has been effective, change must be documented in the target outcome variable. Therefore, baseline data must be collected prior to initiation of intervention. Baseline data provides a comparison point in which to judge your intervention data. Data should be examined for a change in trend, level, variability, and immediacy to establish effectiveness. Academic interventions may warrant a longer intervention period (e.g., 6 weeks) before determining effectiveness whereas you may be able to make decisions regarding effectiveness of behavioral interventions more quickly.

You should first ask yourself if the target behavior is so significant that it warrants immediate intervention (i.e., self-injurious behavior). If not, it is essential to collect baseline data to provide a comparison point by which to judge the effectiveness of the intervention through the change in the outcome variable. Rushing to begin an intervention without baseline data may increase the likelihood of retaining an ineffective intervention.

To determine if an intervention appears to be working or not, it is necessary to establish and utilize a goal line. A goal line can be based off of norm-referenced data or team based decisions of where a behavior should be given a specified period of time. For low stakes decision-making, a three-point decision rule can be used to interpret need for change in an intervention. If the least three intervention data points fall below the goal line (when an increase is desired), it may be necessary to make a change in the duration, type, or frequency of the intervention.

For a student who is identified as at-risk, progress monitoring should occur at a minimum weekly but more frequently when possible. The frequency in monitoring may depend on the severity of the behavior of concern. For example, if a student is slightly delayed in reading and attends a small group reading intervention twice weekly, progressing monitoring data may be collected 1-2 times weekly. If a student is engaging in a more severe behavior such as frequently elopement or self-injurious behavior, behavior may be progress monitored more frequently as the intervention may be more frequent in nature. Feasibility should also be considered when progress monitoring.

There is no hard and fast rule regarding the total number of data points necessary to make effective decisions. The more data per phase collected, the more apparent the behavior pattern is. Generally speaking, enough data must be collected per phase in order to summarize the nature or pattern of the behavior. Some argue that 6-8 data points are needed to make a determination regarding effectiveness of an intervention.

One should first consider if the use of an effect size is necessary. Effect sizes provide information regarding the internal validity of an intervention and may be necessary in high stakes decision-making or Tier 3 cases. If effect sizes are appropriate, percent of non-overlapping data (PND) and percent of all non-overlapping data (PAND) are common and easily understood effect size techniques that can be used, however, empirical interpretation of data should focus on estimates of slope value.

The duration of an intervention largely depends on the type of intervention being implemented. Academic interventions often last longer in duration as academic skills are acquired over weeks. Behavioral interventions can often elicit immediate changes in behavior and therefore intervention can be shorter in duration. Intervention length should be determined by effectiveness and growth according to goal lines. Intervention should be constantly monitored for effectiveness and altered to account for progress towards goal and lack of progress.