Quizzes 7 & 8 (PBR 9-10 SSR 7-8)

Delayed Multiple Baseline Design

Delayed Multiple Baseline Design • Used when a withdrawal design no longer is possible • OR when other behaviors, settings, or individuals emerge that are in need of intervention • Baselines are not measured simultaneously • Potential problems in demonstrating a functional relationship • Advantages: • May allow the use of fewer resources • May allow the researcher to extend the study to new behaviors, settings, and individuals that had not been target priority The major advantages to the delayed multiple baseline design is that it may allow the use of fewer resources and may allow the researcher to extend the study to new behaviors, settings, and individuals that had not been targeted a priori. Cooper et al. (2007) noted three limitations to delayed designs. First, delaying treatment for other behaviors, in other settings, or with other individuals may be problematic, although this difficulty is inherent in multiple baseline designs in general. Second, fewer data points may have been gathered and the length of the various baselines may differ. Third, the use of delayed baselines when new behaviors, settings, or individuals emerge may mask the effects of the independent variable on the dependent variable.

Prediction, Verification, and Replication in Changing Conditions Designs

Depends on the specific combinations of treatments and when and where baseline conditions are reintroduced • Prediction • Ex.: You could predict that the "A" condition will change as a result of the introduction of "B" and "C" • Verification • Depends on the return to baseline following a specific treatment condition • Replication • Occurs when the conditions are reintroduced The prediction, verification, and replication of changing conditions designs, in general, depend on the specific combinations of treatments and whether and where baseline conditions are reintroduced. For example, in an A-B-C design, the prediction criterion can be met but the verification criterion can only partially be met. In other words, you can predict that the A condition will change as a result of the introduction of B and C, and verification is partially met when the pattern changes. However, the other prediction—that the pattern would remain the same given a return to baseline—cannot be verified. The verification of other combinations depends on the return to baseline following a specific treatment condition. So, for an A-B-A-C-A-D design, verification for both the B and C conditions (but not D) could be established; replication occurs when the treatment condition(s) are re-introduced.

example of probability sampling

If you know that 100 people called your anti-bias hotline last month and you want to review 20 of the case records to see what the presenting issues were, then each case would have a known (20 out of 100 or 1 in 5) chance of being selected. In addition, you might have information about gender, age, reason for call, etc. to which you can compare those in your sample to see how representative your sample is.

Establishing a Historical Imperative for Protection of Human Subjects

In 1974, the National Research Act established institutional review boards and also the National Commission for the Protection of Human Subjects in Biomedical and Behavioral Research. The National Commission was formed to create a set of guiding principles for researchers, and their work resulted in the Belmont Report

Non-Probability Sampling

In non-probability sampling, it is not possible to estimate the person's, or case's or study unit's chances of being included in the sample because you do not know how big the population is that the sample is being drawn from. Likewise, you have no empirical data about the distribution of key variables in the universe. This is quite often the case in social work as we regularly deal with populations of an unknown size and unknown characteristics. When you use a non-probability sampling strategy for your PBR study, you do not try to make inferences about how the findings apply to other people in the universe; rather, you acknowledge that the results of the study only reflect the sample that you studied. This still provides important practice and policy data related to your sample.

Sample Size

In probability samples, the researcher can calculate the size of the sample that is needed in order to know how confident she or he can be about the results. The goal is to study a practically feasible number of participants while minimizing sampling error.

Probability Sampling

In probability sampling you can calculate each person's chance of being included in the sample and the likelihood that the generalization drawn from the sample is valid. To employ probability sampling, you must know the size of the entire population to which the question applies, and how many units you have the resources to sample. Ideally, you would also know the distribution of key variables in the sample universe as well. Even when you know little about the distribution of key variables in the universe, when you use a probability sample for your PBR study and the sample is large enough, the results that you get can be safely generalized. In that instance, you make an assumption that the things you found out about your sample also apply to the larger population. Of the two basic types of sampling, probability sampling is the more rigorous from a research point of view and provides the safest inferences as well as a measure of how safe those inferences are. It's what we refer to as statistical significance which essentially tells you how safe it is to generalize from a sample to a total population.

So what are Ethics Anyway?

In the US the National Association of Social Workers (NASW) describes the purpose of codes of ethics as to "set forth values, ethical principles, and ethical standards to which professionals aspire and by which their actions can be judged" (NASW, 2008, p.4). Accordingly, the NASW code of ethics outlines six underlying values that serve as the foundation for the code of ethics, namely service, social justice, dignity and worth of the person, importance of human relationships, integrity, and competence (NASW, 2008).

PBR Sampling

It involves studying a sub-set of units of analysis (e.g. individuals, case records, programs, etc.) and generalizing the findings to a larger population from which the sample was drawn. Your study sample refers to the people or cases that you are gathering data about or from. Sampling eliminates the worry about whether it is safe to generalize from your findings to the larger population from which your sample was drawn.

Covariance Among Dependent Variables

It is also important to note the relationship among the dependent variables. Realizing that variables in the treatment may covary, it is important to select dependent variables that exhibit some degree of independence (Tawney & Gast, 1984). In our example, if a student was likely to increase appropriate comments in all three classes when the independent variable is introduced only in the first class, then the dependent variables would covary (or change at the same time and in the same direction even in the absence of the intervention in classes two and three) (Datilo, Gast, Loy, & Malley, 2000). These variables would not be considered sufficiently independent of one another for research purposes because one could not achieve the desired replication of effects across the dependent variables (see Figure 8-3 for an example of covariance among dependent variables). This independence is important in maintaining experimental control, yet selecting independent variables that are completely unrelated would also be undesirable. A balance between these two areas of concern will produce the best choice. The researcher should select dependent variables (individuals with the same behavior, different behaviors in the same individual, or different settings in which the same individual exhibits the same behavior) that are functionally similar so that they would likely change similarly in response to the same treatment, while at the same time not be likely to change until that treatment is specifically introduced to the particular dependent variable (e.g., -ed endings and spelling changes may very well respond similarly to instruction that improves recognition of -s endings but would be unlikely to change until the instruction is delivered specifically for each of those behaviors). Tawney and Gast (1984) referred to these dependent variables (or baselines) as being at once functionally similar and functionally independent of one another. Again, these dependent variables must be measurable using the same method for recording behavior as well so that results are easily compared from one dependent variable to the next. Also, the dependent variables should be measured concurrently, and the possible influences of other variables should either be reasonably equal or their influence controlled for each dependent variable

Application Practice: Derrick

Limitations of the Study • Covariance that occurred during the lunch period confounds the prediction • Limited follow-up in the study • Home environment was not addressed, even though it was an important component for Derrick social conversation

Prediction, Verification, and Replication For Multiple Baseline Designs

Prediction • After baseline data are stable, the prediction would be that there would be no change in the data path for the dependent variables if there was no intervention effect • Verification • When the intervention is implemented, the data path changes predictably for the dependent variable • Verification is evident if the data path changes in a predictable manner through a phase change, as from baseline to intervention • Replication • The prediction and verification are repeated for each dependent variable • The form of replication provides a convincing argument for the presence of a functional relationship between the dependent and independent variables Beginning the collection of baseline data simultaneously across all dependent variables adds an important feature to multiple baseline designs. It allows the researcher to infer a verification of the prediction that the baseline behaviors would have remained stable and unchanged if the intervention had not been implemented (Kucera & Axelrod, 1995). As discussed in Chapter 4 (and here, in a modified discussion of Tawney & Gast, 1984), verification is evident if the data path changes in a predictable manner through a phase change, as from baseline to intervention. In other words, if the data path line remains constant across the baseline and intervention phases, then the independent variable yields no effect on the dependent variable (i.e., no change in the target behavior when the intervention is implemented) and there is no verification. If there is a change in the data path, then the possibility exists that the independent variable is responsible for the change and there is verification that the data path has changed with the introduction of the treatment. In a multiple baseline design, replication of this prediction and verification may occur when the data paths of subsequent dependent variables follow patterns similar to the first (and subsequent) variables. It is important to note the importance of maintaining control of the extraneous variables that have the potential to influence results across variables. The only changes that should be evident are found in (a)the dependent variable (and, again, those may be different behaviors of the same individual, the same behavior in the same individual in different settings, or the same behavior in the same setting with different individuals) and (b)the treatment, or independent variable.

Power Analysis

A statistical procedure called power analysis tells us what sample size is necessary for us to test cause—effect hypotheses with a desired degree of confidence.

Advantages and Disadvantages of the Changing Criterion Design

Advantages • Helpful when terminal goal takes a relatively long time to reach and is challenging • Treatment does not have to be withdrawn to show its functional relationship with target behavior • Disadvantages • Behavior should change only to specified criterion level, but this is not always educationally desirable • Requires a lot of planning

Mechanics of the Multiple Baseline Design

After baseline data have been obtained for all three, the researcher implements the intervention for the first dependent variable while maintaining baseline conditions for the other two. When the criterion is obtained on the first behavior, in the first setting, or with the first individual subject following intervention, the intervention may be implemented and analyzed as to its effect on the second dependent variable. Meanwhile, baseline conditions are maintained with the third dependent variable.

The Changing Conditions Design

An extension of A-B design -Various changing conditions designs can be used to evaluate the effects of more than one treatment or a combination of treatments • The simplest changing conditions design is A-B-C -A represents baseline -B represents Treatment 1 -C represents Treatment 2 -The hyphen between Treatments 1 and 2 represents that they are presented independently

Placement of the Subphases

As discussed above, the changing criterion design is used to systematically increase or decrease a target behavior toward a terminal goal. Experimental control is dependent, in part, on the variation of the number and length of the subphases and the magnitude of the criterion levels. However, if the direction of the criterion changes is always the same, it is more difficult to demonstrate that the changes "are not naturally occurring due to either historical, maturational or measurement factors"

Length of Each Subphase

As noted previously, in the changing criterion design, the level of responding of each subphase actually serves as a baseline for the subsequent subphase. It is therefore important that each subphase continue until stable responding has occurred. The nature of the design, however, should generally allow this to occur quickly. In other words, the presentation of a new criterion level should result in an almost immediate change in the target behavior to that new level. The evidence for the functional relationship between the independent and dependent variables is strengthened when the behavior changes to exactly the criterion level and stays at that rate until the criterion level changes in the next subphase. For this reason, the actual length of the subphase should vary to demonstrate that control.

Multiple treatment designs

Changing condition designs that include the reintroduction of baseline conditions are frequently referred to as multiple treatment designs. Note that the terms treatment and intervention (or independent variable) are used interchangeably, and the reader should not be confused when reading these various terms in the research literature.

Prediction, Verification, and Replication in the Changing Criterion Design

Prediction • Prediction of the levels of future behaviors is made when stable responding is attained within each subphase • Verification • Is possible when either of two of the previously discussed suggestions to increase internal validity is made • By varying the lengths of the subphases • Demonstrated when the direction of the criterion levels is reversed and the behavior returns to a previously set criterion level • Replication • Occurs every time that the behavior changes in the predicted direction based on the predetermined criterion levels Although prediction and replication are easily addressed within a changing criterion design, verification is somewhat more difficult to demonstrate (Cooper et al., 2007). In general, prediction of the levels of future behaviors is made when stable responding is attained within each subphase. Verification is possible when either of two of the previously discussed suggestions to increase internal validity is made. By varying the lengths of the subphases, verification of the treatment effects is made. Similarly, and perhaps more convincingly, verification is demonstrated when the direction of the criterion levels is reversed and the behavior returns to a previously set criterion level. Replication occurs every time the behavior changes in the predicted direction based on the predetermined criterion levels.

Types of Sampling

Probability Non-probability

Sampling Error

Refers to the difference between the responses of the total population and the answers of your sample. The bigger your study sample, the better the chance that it accurately reflects the group that you are sampling from (population), because a greater proportion of the population is included. The more closely the participants' scores match those of the total population, the smaller the sampling error. It should be noted that the increase in sample size is not proportionate to the increase in accuracy.

A-B-C Designs and Response to Intervention

Response to Intervention (RTI) model ▪ Widely used in general education and special education to help identify students with learning and behavior problems A tiered approach in which students progress through stages of more intense intervention with the goal of remediating problems ▪ Goal of avoiding the need for a referral for special education Using the A-B-C design with the RTI model ▪ Each tier can be thought of as a different letter It might seem logical that the A-B-C design could be used to see which treatment (B or C) is more effective. However, this is not possible for at least two reasons. First, the changes between B and C might be due to other factors (as in an A-B design). For example, a teacher might conclude that technique C is superior to technique B because a student performed much better in the C condition. However, the improvement in C might be due to any of a number of factors, such as maturation or a new medication. Second, there could be a sequencing or a cumulative effect.

The A-B-A-C (multiple treatment) Design

The difference between A-B-C design and A-B-A-C design • InA-B-A-C,anotherbaselineconditionisintroducedbetweenthetreatments • The advantage of A-B-A-C design • The functional relationship can be better established because of the return to the baseline before the "C" condition There are an endless number of adaptations of this design. For example, the experimenter could evaluate the effects of three interventions (A-B-A-C-A-D) or even more. Again, note that the potential for verification is present but the potential for replication is not because no specific treatment is evaluated more than once. It is also possible that the experimenter might combine treatments. For example, in an A-B-A-C-A-BC-A-BC design, A phases would be baseline, the B phase might be the use of prompts, the C phase the use of praise, and the BC phase a combination of prompts and praise. In these types of designs, it is important to be aware of which treatments or treatment combinations are being evaluated in isolation as well as the order in which they are presented in order to avoid making false interpretations. As mentioned, the various possible phases that can be evaluated in a multiple treatment design are virtually limitless. However, the complexity and length of a study will depend on the number of treatments to evaluate and the degree of experimental control that is desired.

Magnitude of Criterion Changes

The magnitude of criterion changes focuses on the question "How much change in the target behavior is required before the subject receives the contingency (intervention)?" This is a very important question—if the required change is small, the subject might progress, but it would be difficult to determine whether the change was not due to other factors, such as maturation or practice effects. If the required change is too large, however, there are at least two possible problems. First, because the target goal will be reached in fewer subphases, there might not be enough subphases to demonstrate experimental control in the study. Second, requiring drastic changes might contradict good instructional practices One logical method of determining the criterion changes is to use the baseline data as an indicator. In general, smaller criterion changes should be used for more stable behaviors, whereas larger changes would be necessary to demonstrate control for behaviors that are more variable (Hartmann & Hall, 1976).

Number of Criterion Changes

The number of criterion changes actually refers to the number of subphases that should be included in the study. The determination will depend on both the length of the subphases and the magnitude of the criterion changes. This demonstrates the interrelationship of these issues.

Advantages and Disadvantages of the Changing Conditions Design

The primary advantage of changing conditions designs is the ability to evaluate the effectiveness of more than one treatment, a modification of a treatment, or a combination of treatments. A primary disadvantage of the A-B-C design, in particular, is the possibility of sequencing effects—that is, the effects of C could be influenced by B. negatives are the issue of the irreversibility of either the behavior or the treatment. The first occurs when the behavior, once changed, is likely to remain even after the treatment is withdrawn. The second occurs when the effects of a treatment are maintained even after it is withdrawn. The sequencing effects noted in the A-B-C design are minimized by the re-introduction of the baseline between treatments but does not eliminate the possibility of these effects. The following are the advantages of changing conditions designs: When the effects of more than one treatment or treatment combinations needs to be determined; For multiple treatment designs: When a clear functional relationship between the independent and dependent variables needs to be demonstrated; When the nature of the target behavior is such that it can be reversed when the treatment is withdrawn; When the nature of the treatment is such that its effects are not present on the target behavior after it is withdrawn; When withdrawal of treatment does not compromise ethics. Conversely, the disadvantages of changing conditions designs are: It doesn't take into account sequencing effects of more than one treatment (especially in the A-B-C design); For multiple treatment designs: Inappropriate when the target behavior is not reversible; Inappropriate when the treatment effects will continue after the treatment is withdrawn; Inappropriate when it is not educationally or clinically desirable for the behavior to return to baseline levels; Inappropriate when the target behavior is such that withdrawal of effective treatment would be unethical (e.g., dangerous behavior).

Multiple Baseline Across Behaviors

The same intervention is applied to similar behaviors in the same individual in the same setting • Advantage: Generality of intervention effects for similar behaviors within the same individual can be demonstrated • Disadvantage: The possibility of covariance among behaviors, which weakens the demonstration of a functional relationship The critical issues in implementing a multiple baseline across behaviors design include: Selection of an individual participant who displays multiple behaviors (at least two but preferably three or more for a convincing argument for a functional relationship) in a single setting; Functional similarity and functional independence of those behaviors as one might be able to determine a priori; A reasonable expectation that the same variables (extraneous or systematic) will exert equal influence on each of the dependent variables; Selection of a treatment or independent variable that can be expected to produce a similar and independent effect on each of the dependent variables; A consistent recording procedure for each of the target behaviors and a criterion level for decision making; and Confidence that the resources and time needed to record multiple baselines and subsequent intervention will be maintained throughout the study.

Multiple Baseline Across Settings

The same intervention is applied to the same individual in different settings • Advantage: Generality of intervention effectiveness with the same individual in different settings may be demonstrated • Disadvantage: Extraneous variables that may influence responding in different settings may be difficult to control or predict The multiple baseline across settings design is similar to the across behaviors design in that only one subject is identified. The researcher identifies two or more settings (again, generally at least three) in which the individual emits the same behavior. The same subject is treated for the same behavior in different settings. For example, a student might be treated for compulsive talking in the classroom, in the hallways, and in the cafeteria. Settings need not be literally interpreted, however, to mean different physical environments. Settings may include functionally similar situations that are still independent of one another. For example, a student's appropriate use of punctuation might be addressed in journal writing, formal written products in a content area, and in writing job application letters. The physical environment may not actually change, although the situations may be different enough that the researcher may not expect the behavior to change in any of those situations until the intervention is applied. Critical issues in the implementation of the multiple baseline across settings design include: Selection of an individual subject who displays the same target behavior in multiple settings; Selection of settings that are functionally similar but also independent of one another as one may best determine a priori; A reasonable expectation that the same variables will be exerting the same influence in each of the settings; Selection of a treatment or independent variable that can be expected to produce similar effects in each setting; A consistent recording procedure for each setting and a criterion level for decision making; and Confidence that the resources and time needed to record data in multiple settings will be maintained throughout the study.

Describe Your Sample

When communicating about your study to others, it is vital that you provide a full description of your sample and how it was drawn, including the extent to which it varies on demographic characteristics that might be relevant to the interpretation of your findings by others, like gender, age, race, ethnicity, socioeconomic status, education level, sexual orientation, and marital status. You should also clearly describe the extent that the sample does or does not mirror the larger population. If a sizable number of individuals refuse to participate or fail to complete the study protocol, you should make every effort to describe the differences between those whose needs, knowledge, attitudes and/or behaviors are represented in the study data and those whose are not.

Calculating the Required Sample Size

You will need three things: ◦ The confidence level that you are applying - the degree of certainty that you want to have that your sample's responses are reflective of their population (e.g., 99, 95, or 90 percent). ◦ The confidence interval that you are willing to tolerate - the range that you are willing to be "off" from the actual responses of the population that you are studying (e.g., + 5 percent). ◦ The size of the population

Advantages and Disadvantages of the Multiple Baseline Design

here are several advantages to the multiple baseline design. First, the withdrawal of an effective treatment is not required to demonstrate the functional relationship between the independent and dependent variables (Baer et al., 1968). Second, the sequential implementation of the independent variable parallels the practice of many teachers (Alberto & Troutman, 2013). Third, generalization of behavior change is monitored through the design. Fourth, the design is easily conceptualized and used (Cooper et al., 2007). A multiple baseline design should be used in the following situations: When withdrawal designs are not feasible because of ethical concerns; When there is more than one target behavior, setting, or individual in need of treatment; When the effects of the independent variable cannot be withdrawn or reversed. Multiple baseline designs also have their disadvantages. These include the possibility of covariance and the aforementioned result that a functional relationship is not clearly demonstrated (Datilo et al., 2000). Verification relies on dependent variable levels not changing until the independent variable is introduced and then changing in a manner similar to that for any previously treated behaviors. Second, the multiple baseline design does yield data related to general effectiveness of the independent variable in treating various behaviors, in different environments, or with various individuals, but our ability to analyze the dependent variable is less than that for other designs because, as a rule, the treatment is applied in only one intervention phase. The withdrawal design can use multiple intervention and baseline phases to clearly demonstrate the functional relationship (Cooper et al., 2007). Finally, implementing a multiple baseline design can be time-consuming and may require substantial resources because two or more dependent variables are being measured simultaneously.

The Changing Criterion Design

• Changing criterion design involves the evaluation of the effects of a treatment on the gradual, systematic increase or decrease of a single target behavior • Accomplished by carefully changing the criterion levels necessary to meet contingencies to increase behavior or to decrease behavior Changing Criterion Steps • Step 1 Carefully define the target behavior • Step 2 Collect baseline data • Step 3 Determine level of performance (criterion levels) • Determine terminal behavior or goal • Determine criterion level for the first subphase • Establish the criterion levels for the subsequent subphases • Step 4 Begin the intervention • Step 5 Introduce the next subphase level after the first criterion level is met • Step 6 Continue through each subphase until the terminal goal is reached

Multiple Probe Design

May be adapted to any of the basic designs • Data probes are taken during baselines rather than continuous measurement • Reduces need for resources • Potentially leads to problems if probes are too infrequent or do not suggest steady baseline recording The multiple probe design (Horner & Baer, 1978) may be used as an adaptation to designs addressing multiple behaviors, settings, or individuals. The primary variation in the multiple probe design is to decrease the collection of data across multiple baselines (and possibly during follow-up or maintenance phases). In this adaptation, the researcher collects data across the multiple baselines at the study's outset, but does not maintain continuous recording of all baseline measures before the introduction of the intervention. Rather, the researcher makes periodic recordings of baseline levels to ensure that no significant changes have occurred before the introduction of the intervention The periodic probes may be used because they reduce the need for resources that may be unavailable to maintain continuous recording of behavior during baseline phases, because baseline measures are causing severe reactivity, or because there is a strong a priori assumption of stability (e.g., the target behavior is not likely to be emitted until the intervention is introduced, as it does not yet exist in the individual's behavioral repertoire; Horner & Baer, 1978). Also, once optimal or criterion responding is achieved during an intervention phase, the researcher may resort to data probes to ensure that changes are being maintained. he major advantage to the multiple probe design is that fewer resources are required, as there is no continuous measurement of multiple baselines. The major disadvantage is that the functional relationship may be more difficult to demonstrate. The researcher may wish to take several precautions. First, it must be ensured that an adequate number of probes are conducted so that one may easily infer that those probes do represent a true depiction of baseline responding. Second, should a probe result in a measurement that significantly deviates from others, the researcher may need to implement continuous (or at least more frequent) recording for that baseline in order to obtain insight into why this may be occurring and to establish a true baseline level of responding (Horner & Baer, 1978). Third, the researcher may wish to conduct a short but continuous baseline measure for each behavior, setting, or individual just before the introduction of the independent variable to assist in establishing a better depiction of baseline level responding.

The Basic Multiple Baseline Design

Multiple Baseline Designs • Demonstrate a functional relationship between the target behavior and intervention by replicating the intervention effects • with two or more behaviors • in two or more settings • with two or more individuals • The three major types of multiple baseline designs are multiple baseline across behaviors, settings, and subjects • The researcher takes repeated measures of baseline performance concurrently on two or more baselines Multiple baseline designs may be the most appropriate single subject designs to use for a variety of reasons (Baer, Wolf, & Risley, 1968). These include the following: (a)when withdrawal or reversal designs may not be feasible because of ethical concerns about withdrawing treatment that is working (Harvey, May, & Kennedy, 2004); or (b)when there are practical considerations, such as more than one person or setting needing interventions; or (c)in cases in which the independent variable (treatment) should not be withdrawn or the achieved target behavior cannot be reversed (e.g., allowing verbal threats to once again increase after a decrease has been obtained, learning to match consonant sounds with the appropriate alphabet letter). Actually, in a multiple baseline across behaviors design, two or more target behaviors of the same individual receive the same treatment in the same setting. In the multiple baseline across settings design, the researcher is applying the same intervention to the target behavior of the same individual in two or more settings. In a multiple baseline across subjects design, the researcher is applying the same intervention to the target behavior of two or more individuals in the same setting. Some experts prefer to conceptualize each baseline (be it different behaviors in the same individual, different settings where the same individual's target behavior is occurring, or different individuals exhibiting the target behavior in the same setting) as a different target behavior or dependent variable.

Types of Probability Sampling

Simple random sampling - the equivalent of pulling numbers out of a hat. ◦ One can use a computer program to generate random numbers or use a random numbers table. Also, many statistical techniques that are employed in quantitative analysis are based on the assumption of simple random sampling. These data-analytic techniques tell us about the strength of relationships (i.e. effect size) as well as how safe it is for us to generalize from the sample to the universe. So sample size as well as how the sample is drawn are important considerations because they may rule out using our most powerful statistical techniques. Dodd, Sarah-Jane. Practice-Based Research in Social Work (p. 119). Taylor and Francis. Kindle Edition. Systematic random sampling - involves selecting every nth case (e.g., 3rd, 5th, or 10th), depending on your sample size and the size of the total population from which your sample is being drawn. Stratified random sampling - requires dividing the population into groups or strata based on a specific characteristic that is central to the study purpose and then draws simple random or systematic random samples from each of the groups created. ◦ Proportionate, stratified random sampling - involves matching the proportions in the larger population. Cluster random sampling (also referred to as multi-level or multi-stage sampling) - is a procedure in which random samples are drawn in stages from a series of naturally occurring groups (clusters). ◦ This approach is often used when the size or spread of the population being studied is so large that it is not feasible to draw a simple or systematic random sample without first partializing the population into smaller more manageable units.

Types of Non-Probability Sampling

Snowball sampling - the researcher starts with a few people who meet the criteria of the study, and then then you ask them to recommend other people who would also be appropriate. Snowball sampling - the researcher starts with a few people who meet the criteria of the study, and then then you ask them to recommend other people who would also be appropriate. Purposive sampling - involves actively seeking subjects that you think will have the specific combinations of characteristics that are relevant to your study.

Example of representativeness

So—if our sample of 20 involved 50 percent men and 50 percent women but the universe of callers involved 25 percent men and 75 percent women, our sample would not be representative and generalizations about hotline "callers" would be suspect. If our study question concerned how men callers and women callers differed in their reasons for calling, this 50/50 distribution would be ideal. However, if our concern was primarily about the kinds of calls and requests for help that came in to the hotline in general, our sample would be problematic because men would be overrepresented and women underrepresented and that might have implications for the types of calls that came in.

Multiple Baseline Across Subjects

The same intervention is applied to the same or similar behaviors, in the same setting, to different individuals • Advantage: Allows researcher to demonstrate the effectiveness of an intervention with more than one individual who displays a similar need for behavior change • Disadvantage: Covariance among subjects may emerge if individuals learn vicariously through the experiences of other subjects • Also, identifying multiple subjects in the same setting who are functionally similar yet independent of one another can be difficult All other variables are held as constant as possible. One should be aware that the phrase same target behavior need not be literally interpreted to mean exactly the same behavior. Such an example might include one student who is disruptive by making noises, one who is disruptive by talking out during instruction, and a third who is disruptive by talking to classmates during instruction. Although these target behaviors of different individuals are not exactly the same, they may be functionally similar yet still independent of one another. The researcher should logically be able to operationally define the target behavior of disrupting class in such a manner that each of the individual responses given would be examples of the target behavior. The critical issues in implementing a multiple baseline across subjects design include: Selection of individual participants who display the same target behavior in the same setting; Selection of individuals who are similar enough to one another to expect each would change his or her behavior in response to the same intervention and yet not likely to change his or her behavior until the intervention is specifically implemented to treat his or her behavior; A reasonable expectation that the same variables will exert the same influence on each of the subjects; Selection of an independent variable that is likely to have a similar effect on each subject; A consistent recording procedure for all subjects' behavior and a criterion level for decision making; and Confidence that the resources will be available to maintain data collection and intervention across the life of the study. Purpose of the study - Bennett, Ramasamy, and Honsberger (2012) • To determine the effects of covert audio coaching (CAC) on the development and maintenance of employment skills of high school students with autism spectrum disorders (ASD) • The design • Multiple baseline across subjects was used • Collected repeated measures of photocopying performance

Representativeness

This is a central objective of sampling and refers to the extent to which the distribution of key variables in the universe is mirrored in the sample. When this is true, then it is safe to generalize findings from the sample to the universe.

Issues Related to Changing Criterion Designs

Three important issues to consider 1. The length of each subphase 2. The magnitude of the criterion changes 3. The number of subphases or criterion changes • Another issue to consider 4. Placement of the subphases