Amego Inc Behavior Analyst Certification Exam Preparation Module
Amego, Inc. Behavior Analyst Certification Exam Preparation Module II Measurement This Module is based on the BACB Task List 4 Section IA. & IH. : Measurement Developed by: Michael Weinberg, Ph. D. , BCBA-D c. 2014 All Rights Reserved; may not be copied or reproduced without permission of Amego, Inc.
Module II. Measurement Behavioral Observation and Data Collection • • • Behavior must be directly observable and measurable Different methods and approaches have been developed for the observation of behavior depending on the type of behavior, the situation, and other factors It is important to be aware of the type of data collection and data display to use for different types of data Also, automated systems of behavior data collection and direct observation methods are used, although most applied studies and intervention uses direct observation by one or more observers Inter-Observer Agreement (IOA): This is a measure of agreement among multiple observers in applied situations to ensure that different people who are trained observers agree with what was taking place and is a critical aspect of measurement. Publications in Applied Behavior Analysis typically require at least some IOA sessions to occur during the study. It is also important for intervention to ensure that there is agreement that the behavior has indeed changed. Know different methods of calculating IOA for different types of data, including continuous measures, time measures, discrete trials, time sampling. Be aware of the IOA calculation method for scored interval, unscored interval and interval-byinterval recording.
Properties of Behavior Direct properties or dimensions of behavior Countability of behavior 1) Repeatability: Behavior can be counted over time. This can be a frequency count, event recording, etc. (continuous or discontinuous measures) Behavior Occurs in Time 1) Temporal Locus: Behavior occurs at a particular point in time. This can be latency or IRT, or time frame in which behavior occurs. 2) Temporal Extent: Behavior occurs during some time period, or measuring duration of behavior Combination of these dimensions 1) Calculating a rate measure includes combining repeatability (frequency) and temporal locus (i. e. rate = frequency/time) 2) Mean IRT, latency, duration (averaging these measures for some observation period)
Module I. Introduction to Applied Behavior Analysis Basic Concepts and Terms Cont’d • • Types of Data NOMINAL: Naming of items or objects, like the no. of cars produced in a given year by Chrysler, Ford, GM , Honda, Toyota, etc. ORDINAL: Rank ordering of data – still nominal data set labels, but are placed in order (e. g. the rank order of the tallest to shortest student in a class, the highest to lowest dollar amount earned by major U. S. auto manufacturers, etc. ) • INTERVAL: Data which can be added and subtracted legitimately, but have no true zero point, so you cannot divide/multiply or use them to talk about percent or proportion (e. g. temperature – Fahrenheit and Celsius; IQ) • RATIO: Can perform all mathematical operations legitimately – addition, subtraction, multiplication and division (because the data have a true zero point) – Egs: no. of cars produced each year, incidents of physical aggression by a teen in a given month, no. of automobile recalls in a given year, no. of yards run by your favorite football team in a given season, etc.
Module I. Introduction to Applied Behavior Analysis Basic Concepts and Terms Cont’d Appropriate graphic displays for types of data • Nominal and Ordinal Data are discrete and are not in any logical continuum (one cannot say that Ford and GM are on a continuum). Hence such data must be plotted using a bar graph. E. g. , the number of correct answers by each student in a class on a math test cannot be displayed as a line graph because the students are individuals (nominal), and not on some continuum like time, height, weight, temperature, and so on. So, these data must be graphically displayed as a bar graph. • Ratio and Interval Data are on a continuum and should be displayed as a line graph (simple frequency curve or cumulative frequency curve, and possibly as a Standard Celeration Chart). One could also plot such data as a histogram, in which the bars are touching each-other at their outer perimeters. The center points of the histogram (top center of the “bar”) can be drawn in and the lines connected to convert it to a simple line graph. The histogram tends to show the area that the data point covers. Behavior data recorded by behavior analysts, such as self-injury, aggression, property destruction, communication skills, word use, social interactions, etc. , are usually interval type data.
Module I. Introduction to Applied Behavior Analysis Basic Concepts and Terms Cont’d Types of data collection Two main categories: Direct & Indirect Methods • Frequency/Rate = (freq. /time) • Duration • Latency • Intensity • Event Recording • Inter-Response Time (IRT) • Permanent Product (Indirect Measures) • Trials to Criterion (Indirect Measure) • Percent (derived measure)
Sample Bar Graph
Sample Histogram from www. analyzemath. com (2014)
Sample Line Graph
Module II: Measurement Data Collection Methods • Continuous measurement such as frequency, duration, latency, Inter-response times (IRT) • Event Recording, such as recording instances of behavior (talk-outs, aggression, SIB, etc. ) during a time period such as in a class period at school each day Use of Time Sampling techniques includes dividing time into short segments, usually 10 or 20 second intervals, and data from direct observations of the behavior of the person(s) being observed are recorded in the boxes for that interval. These data are often converted into percentage of recorded intervals with occurrence or non-occurrence of the behavior a) Typically, a “+” indicates occurrence of the behavior (scored interval) and a “-” indicates that the behavior did not occur during that interval (unscored interval), depending on the specific time sampling method used b)Convert scored intervals into percent intervals (divide scored intervals by total no. of intervals recorded for a particular observation time frame) •
Module II: Measurement • There are Three (3) major types of Time Sampling data recording methods: 1) Partial Interval Recording- behavior occurs at least once during the interval 2) Whole Interval Recording – behavior must be observed for the entire interval 3) Momentary Time Sampling – behavior is observed at the end of the interval. The observer may be prompted to observe and record data at the correct time using a timer device with an auditory or visual cue.
Measurement �Recommendations for use of time sampling data recording: a) For low to moderate frequency behavior, use partial interval recording such which could be talk-outs in class, physical aggression, or any other countable, discrete events b) For high frequency behavior or behavior that takes place in time (high frequency SIB, or “tantrums, ” time on task or engaging in social interactions), use whole interval recording.
Inter-Observer Agreement (IOA) �To increase reliability of data collection, conduct IOA observations. There are essentially two main formulae with some variations. 1) for continuous measures such as frequency, duration, latency, etc. the formula is: IOA= lower number (frequency, duration, latency, etc. ) X 100 higher number (frequency, duration, latency, etc. )
IOA Continued 2) For Trials-to-Criterion or Time Sampling, the basic formula is: IOA = (Agreements/Agreements + Disagreements) X 100 � Special Case formulas: a) For interval x interval, use ALL recording intervals in the denominator and include agreements of BOTH scored and unscored intervals. For example, if you use partial interval recording using 10 intervals, you would calculate IOA by adding the intervals in which both observers agreed on the occurrence (scored interval) and non-occurrence (unscored interval) of the behavior and divide by the total number of intervals (i. e. 10), and multiply by 100 to obtain a percentage.
IOA Continued b) Scored Interval IOA calculation: For this calculation, add the number of instances the observers agreed on occurrence (scored interval) of the behavior and divide by the total number of intervals of agreement of occurrence (scored interval) of the behavior plus intervals in which AT LEAST ONE observer recorded an occurrence (scored intervals), or disagreements, and multiply this result by 100. All instances of agreement of nonoccurrence (unscored intervals) are not used in the denominator. c) Unscored Interval IOA calculation: For this calculation, add the number of instances the observers agreed on nonoccurrence (unscored interval) of the behavior and divide by the total number of intervals of agreement of nonoccurrence (unscored interval) of the behavior plus intervals in which AT LEAST ONE observer recorded nonoccurrence (unscored intervals), disagreements, and multiply this result by 100. All instances of agreement of occurrence (scored intervals) are not used in the denominator.
Evaluating Data �Data are evaluated for treatment or research decisions based upon visual analysis of trend, level and variability, usually on a line graph that plots frequency or rate of the behavior across time. �Trend is the direction of the data path (or slope of the line) �Level is the amount of the behavior for a given data set �Variability refers to how much the behavior goes up and down across observations represented by data points and data path on a graph
Module II: Measurement This Module reviews: �Types of Measurement used in Applied Behavior Analysis �Types of Data �Graphing Conventions �Types of Graphs and When to Use Each �Inter-Observer Agreement (IOA) This module corresponds to Amego, Inc. Probes 3 & 4
- Slides: 17