CDIS 564 • Methods Reference

Data Collection

How you define, measure, and record the dependent variable across your study.

Data collection is the part of your Methods chapter that tells a reader exactly what you measured and how. Everything downstream, including reliability, fidelity, and your results, depends on getting this section precise. Expand each panel below for the detail you need while you draft.
1 Operational definitions ▼

An operational definition states the target behavior in observable, countable terms, so that two people watching the same session would record it the same way. Write the definition before you collect a single data point. If a behavior cannot be seen or heard and counted from your definition alone, the definition is not finished.

A complete operational definition usually names three things: what the behavior looks like (a description a second observer could apply), what counts as one instance (the boundary that separates one occurrence from the next), and what does not count (near-misses and look-alikes you will exclude).

Example from a single case study In a worked narrative-intervention study in this course, each child utterance was first coded as relevant or not relevant before anything was counted. An utterance was relevant if it was about the book and the page the child was viewing, and not relevant if it referred to something off the page. Only then were relevant utterances scored for story grammar elements and mean length of utterance. The definition of what counted as a countable utterance, including how spoken words and words produced on the speech device were each recorded, was written out in a standard operating procedure before coding began.
Example from Namasivayam and colleagues (2024) The authors defined dose as the number of productions the child attempted for each goal. That single operational definition made an otherwise vague idea, how much practice a child got, into something a coder could count from video.
Watch out A definition that uses judgment words such as "engaged," "on task," or "correct" without spelling out what those look like will not survive a reliability check. Turn every judgment word into a description of what the observer would see or hear.
2 Choosing a measurement method ▼

Match the measurement method to the dimension of behavior your research question cares about. The common methods in single case design are below, with the situation each one fits.

MethodWhat it capturesUse it when
Event or frequency How often a behavior happens (a count, or a rate per unit of time) The behavior has a clear start and stop and is brief, for example number of correct productions or number of requests
Duration How long a behavior lasts The length of the behavior matters more than how often it starts, for example time engaged in an activity
Latency How long from a cue to the start of the behavior You are timing responsiveness to a prompt, for example seconds between a question and a spoken reply
Interval recording Whether the behavior occurred within set time blocks (partial, whole, or momentary time sampling) The behavior is continuous or hard to count as discrete events, so you sample it across intervals
Permanent product A lasting result the behavior leaves behind The behavior produces something you can score later, for example a written sample, a recording, or completed work

Most CDIS 564 proposals use event or frequency counts scored from a permanent product, usually a video or audio recording of each session. Recording the session lets you score carefully, rescore for reliability, and confirm fidelity, all from the same file.

Example The narrative study scored story grammar elements and mean length of utterance (frequency-type counts) from video recordings of every session. Namasivayam and colleagues (2024) scored percent consonants correct and whole word match from audio and video, and measured intelligibility by having naive listeners identify target words from recordings.
3 The measurement schedule ▼

Single case design rests on repeated measurement. You measure the same dependent variable the same way, many times, across every phase of the study. The pattern of those repeated measures is what lets you argue the intervention, and not something else, produced the change.

State your schedule explicitly in the Methods chapter: how often you probe, in which phases, and how many data points you need before moving on.

  • Baseline (A): Collect repeated measures until the data are stable, and plan for enough points to establish a predictable pattern. What Works Clearinghouse standards for single case design call for at least five data points per phase, and a common rule of thumb is a minimum of five baseline probes before starting intervention.
  • Intervention (B): Keep measuring on the same schedule so you can see level, trend, and variability change against baseline.
  • Maintenance: Probe again after intervention ends to see whether the change holds over time.
  • Generalization: Probe with new materials, settings, or people to see whether the change transfers.
Example schedules In the narrative study, probes were run weekly or biweekly through baseline and intervention, with at least five baseline probes, and three follow-up sessions in the week after the last intervention session, each including one maintenance probe and one generalization probe. In Namasivayam and colleagues (2024), dose was calculated from 30% of the intervention sessions, with each session split into five-minute windows and three windows chosen at random for the count.
Watch out In a multiple baseline or multiple probe design, the schedule includes a staggered start. Intervention begins for the next tier only after the first tier shows a stable change, for example three consecutive data points that do not overlap with its baseline. Build that decision rule into your schedule.
4 Who records, and on what form ▼

Name the person who records the data and the exact form they use. Consistency here is what makes your data trustworthy. Two ideas do most of the work: standardize the procedure, and give the recorder a data sheet that leaves nothing to memory.

  • Build a data sheet keyed to your operational definitions, so recording is a matter of marking what is defined, not deciding in the moment what counts.
  • Standardize the collection procedure so every session is run the same way. A short session checklist keeps setup, recording, and file handling identical from session to session.
  • If you score from recordings, state the recording setup (camera placement, audio device) and the file naming and storage plan.
Example of a data collection checklist Wence and colleagues (2024) used a video collection checklist so that every session was captured the same way. It walked the person through a private, quiet setup, starting the recording, giving families the same one-sentence description, confirming which book and communication device each session called for, instructing families to read as they normally would, then stopping the recording and saving the file under a set naming convention. A checklist like this is a data collection tool, not a fidelity tool: it standardizes how the measurement session is produced.
5 Collecting across all phases ▼

The same dependent variable is measured with the same method and the same operational definitions in every phase. Changing how you measure between baseline and intervention would make the phases uncomparable, which is the whole comparison your design depends on.

  • Baseline and intervention measurement procedures are identical except for the presence of the intervention itself.
  • Maintenance and generalization probes reuse the baseline measurement procedure.
  • Record enough at each phase to see the pattern, not just a single snapshot.
Example In the narrative study, baseline, intervention, maintenance, and generalization were all recorded on video using the same measurement portion of the session, so productions could be transcribed and scored identically across phases.
6 What to write in your Methods chapter ▼

Your data collection section should let a reader repeat your measurement exactly. Include each of these:

  • The operational definition of every dependent variable.
  • The measurement method for each variable, and why it fits.
  • The measurement schedule across baseline, intervention, maintenance, and generalization.
  • Who recorded the data, on what form, and how sessions were captured and stored.
  • Any decision rules, such as when intervention starts or how many points a phase needs.

The SCRIBE reporting guideline, included in this course, is a useful final check: it lists the measurement and procedure details reviewers expect a single case report to state.

□ Sources ▼