barnes

. The IV is always a nominal (grouping) variable with two or more conditions (groups).

, quasi-experiments and experiments

. There are two types of chi-square tests;

the chi-square goodness of fit is appropriate for studies with one nominal variable, and the chi-square test for independence is appropriate for studies with two nominal variables.

1. What are the benefits of using a nonparametric test?

the computations are much simpler for nonparametric statistics and allow a quick check on trends in the data. Although nonparametric statistics are less powerful, they do provide useful information about differences between groups or relationships between variables and should be included in your toolbox of potential statistical tests.

. The ______ phase is used to determine patterns of the dependent variable in the presence of a manipulation.

. manipulation

Stable baseline

: A baseline that displays no trend (or slope) and little variability and therefore allows for prediction of future behavior.

Spearman's rho (rs ):

A commonly used nonparametric statistic that analyzes the relationship or correlation between two ordinal variables.

Grounded theory: and how does one do it

A method to build theory from data. A researcher applying grounded theory would start with a single case and attempt to understand the case in its entirety, identifying categories and concepts that emerge from the case and making connections between them. Additional cases are collected based on the concepts that emerge, and the concepts are fine-tuned by constantly comparing them across cases until a theory emerges that captures all the cases

Kruskal-Wallis H test:

A nonparametric test used to analyze ordinal data from a study with one variable with at least three levels.

Chi-square goodness of fit:

A nonparametric test used with one nominal variable having two or more categories; tests whether the observed frequencies of the categories reflect the expected population frequencies.

Chi-square test for independence:

A nonparametric test used with two nominal variables having two or more categories; tests whether the frequency distributions of two variables are independent.

Visual inspection:

A nonstatistical technique in which patterns of the A and B phases are compared.

demonstrates the strongest evidence for a manipulation effect

A phase B trend that is in the opposite direction of the baseline trend

AB design:

A simple comparison of the baseline (A) and manipulation (B) phases.

Multiple-manipulation design:

A single N design in which the researcher introduces two or more manipulations over the course of the study.

4. If you are examining the relationship between two variables, and the variables are ordinal, you should conduct a ______.

B. Spearman's rho

2. If your goal is prediction or explanation, but it is not ethical to manipulate an IV, the best choice of designs is ______.

B. correlational design

We compute a Spearman rho when we want to correlate ______.

B. two variables of ordinal data

1. Experiments have higher internal validity than quasi-experiments because they ______.

B. utilize random assignment to IV conditions

can have more or less internal validity based on how well controlled the study is (or how well the researcher minimizes the threats to internal validity

Both quasi-experiments and experiments

. The primary goal of sample-based studies is to ______.

C. extrapolate the results of the sample to the population, the larger the sample, the more likely we can make these extrapolations to the population.

. A single N design is best grouped with ______.

C. quasi-experiments

have the advantage of better external validity because the researcher is not trying to systematically control participants' (or animal subjects') environments.

Descriptive and correlational designs

2. What is a case study?

Detailed investigation of a single individual, group, organization, or event.

utilize random assignment, and therefore have greater internal validity than quasi-experiments.

Experiments

Manipulation (phase B):

In a single N design, repeated assessment of the dependent variable during the implementation of a manipulation (e.g., treatment).

Baseline (phase A):

In a single N design, repeated assessment of the dependent variable in the absence of any manipulation.

1. When is it appropriate to use nonparametric statistics?

In cases where a study violates the assumptions for parametric statistics, it is most appropriate to use nonparametric statistics. data that are most appropriate for a study are nominal (categories or groups) or ordinal (ranks)? Or what if we collect both interval/ratio data and nominal or ordinal data within the same study. Or what if the interval or ratio data in our study violate other assumptions for a parametric test such as a normal distribution or homogeneity of variance? Your study may include data that are on a nominal or ordinal scale. Sometimes the sample data are dramatically skewed so nonparametric statistics are more appropriate. Or you may have a small sample for a pilot study and want to determine whether there are any trends to explore further

What are the assumptions of parametric statistics?

Interval or ratio data Normally distributed variable Homogeneity of variance for groups

Embedded case study:

Investigation of single cases that comprise a group or organization in order to understand that group or organization as a whole.

Chi-square tests (c 2 ):

Nonparametric tests used with nominal data that compare expected versus observed frequencies.

There are two primary reasons for conducting a factorial study.

One is that you have reason to expect that the relationship between your variables will depend on a third, moderating variable. The other reason is when you want to systematically control confounds or extraneous variables.

utilize already existing groups, and therefore have greater external validity than experiments.

Quasi-experiments

. these have the advantage of better internal validity.

Quasi-experiments and experiments include a manipulation of an independent variable (IV)

is one way that a researcher conducting a single N study attempts to rule out alternative explanations for causality

Repeated assessments of the dependent variable

1. What are the differences between parametric and nonparametric statistics?

Shape of the distribution: Parametric statistics assume that the sample data are normally distributed while nonparametric do not. For this reason, nonparametric statistics are sometimes referred to as distribution-free statistics. When it is clear that the distribution of data in a study is significantly skewed, nonparametric statistics should be computed to test the hypotheses. Sample size: Parametric statistics typically have at least 10 participants in each group while nonparametric statistics can be computed with a smaller N where it is difficult to obtain a normal distribution. Some nonparametric tests, however, require at least 5 in a group, and some studies employing nonparametric statistics have large samples. A researcher who conducts a pilot study with a small sample may use nonparametric statistics to check for a trend before collecting a larger sample of data that will be analyzed using parametric statistics. Scale of measurement: Parametric statistics are used with interval or ratio data while nonparametric statistics are used with nominal or ordinal data. As will be discussed later, interval and ratio data can be transformed to ordinal data. This transformation is appropriate when the distribution is not normal or the sample size is too small to perform a valid test using parametric statistics. Homogeneity of variance: Parametric statistics typically assume that the variances in each group are the same. You learned in Chapter 10 that a more stringent t test is used when the assumption of homogeneity is violated in an independent two-group design, and that the one-way ANOVA is a more robust test and can better handle some difference in variances among the groups. Nonparametric statistics make no assumptions about the variances across groups and should be used when samples have very different variances. Interactions: Parametric statistics can test for interactions (such as in factorial designs) between variables while nonparametric statistics test for independence between variables but not for interactions. Power: Parametric tests are more powerful (have a greater probability of correctly rejecting a false null hypothesis) than nonparametric statistics. As just outlined, several assumptions should be met in order to compute parametric statistics in hypothesis testing. A study that meets all of these assumptions is somewhat "protected" from Type I errors because of the care to ensure that only in those cases where the data meet stringent requirements and show very different results for the groups, will the null hypothesis be rejected and the groups assumed to come from a different population than that of no differences between the groups. Computations: The computations for parametric statistics are much more complicated than those for nonparametric statistics. Because of the ease of computation, nonparametric statistics are sometimes used as a quick check on a trend before all the data are entered into a file or during data collection.

how do Parametric and nonparametric statistics differ

Shape of the distribution: Parametric statistics assume that the sample data are normally distributed while nonparametric do not. For this reason, nonparametric statistics are sometimes referred to as distribution-free statistics. When it is clear that the distribution of data in a study is significantly skewed, nonparametric statistics should be computed to test the hypotheses. Sample size: Parametric statistics typically have at least 10 participants in each group while nonparametric statistics can be computed with a smaller N where it is difficult to obtain a normal distribution. Some nonparametric tests, however, require at least 5 in a group, and some studies employing nonparametric statistics have large samples. A researcher who conducts a pilot study with a small sample may use nonparametric statistics to check for a trend before collecting a larger sample of data that will be analyzed using parametric statistics. Scale of measurement: Parametric statistics are used with interval or ratio data while nonparametric statistics are used with nominal or ordinal data. As will be discussed later, interval and ratio data can be transformed to ordinal data. This transformation is appropriate when the distribution is not normal or the sample size is too small to perform a valid test using parametric statistics. Homogeneity of variance: Parametric statistics typically assume that the variances in each group are the same. You learned in Chapter 10 that a more stringent t test is used when the assumption of homogeneity is violated in an independent two-group design, and that the one-way ANOVA is a more robust test and can better handle some difference in variances among the groups. Nonparametric statistics make no assumptions about the variances across groups and should be used when samples have very different variances. Interactions: Parametric statistics can test for interactions (such as in factorial designs) between variables while nonparametric statistics test for independence between variables but not for interactions. Power: Parametric tests are more powerful (have a greater probability of correctly rejecting a false null hypothesis) than nonparametric statistics. As just outlined, several assumptions should be met in order to compute parametric statistics in hypothesis testing. A study that meets all of these assumptions is somewhat "protected" from Type I errors because of the care to ensure that only in those cases where the data meet stringent requirements and show very different results for the groups, will the null hypothesis be rejected and the groups assumed to come from a different population than that of no differences between the groups. Computations: The computations for parametric statistics are much more complicated than those for nonparametric statistics. Because of the ease of computation, nonparametric statistics are sometimes used as a quick check on a trend before all the data are entered into a file or during data collection.

Parametric statistics:

Statistics used to analyze interval and ratio data and that assume a normal distribution and homogeneity of variance between groups.

Nonparametric statistics:

Statistics used to analyze nominal and ordinal (ranked) data or used when the assumptions of parametric statistics are violated.

1. In general, what are the key factors in selecting the appropriate statistical analyses for a study?

The analysis is based on how many variables you are examining and the type of data you have (nominal [or groups], ordinal, interval, or ratio) within the design, purpose, how many variables, scale of measurement

Strengths and Limitations of Single N Designs

The potential to identify a cause-and-effect relationship within a single case is the greatest strength of the single N design (although its ability to live up to that potential depends on how well controlled the design is). good choice for clinicians working with individuals or small groups. Single N designs can also be valuable supplements to randomized-group experiments The repeated assessment also allows for a lot of flexibility., the repeated assessment requires a considerable amount of time and effort. T

ABA reversal design:

The simplest type of reversal design that involves an initial baseline (A), manipulation (B), and a return to baseline (A).

4. What are the strengths and limitations of a case study?

The holistic nature of the case study is one of its greatest strengths. Rather than reducing information into data that can be quantified and analyzed, the case study utilizes primarily qualitative methods to capture the wholeness of a case. However, this holistic nature leads to several criticisms of the case study. First, the case study often relies on anecdotal information that is difficult to verify and is subject to the interpretation of the researcher. Second, the case study lacks control and therefore has limited ability to determine causal relationships. Finally, the details of a case study can be so persuasive that it might bias the public, and even researchers, to weigh results from a single case more heavily than results from other research.

Multiple-baseline across behaviors:

The manipulation is introduced at different times across two or more behaviors.

Multiple-baseline design:

The manipulation is introduced at different times across two or more persons, settings, or behaviors.

Multiple-baseline across persons:

The manipulation is introduced at different times across two or more persons.

Multiple-baseline across settings:

The manipulation is introduced at different times across two or more settings.

Reversal:

The manipulation is removed and the individual returns to a baseline phase.

Some reasons to choose a multilevel design:

The variable you are examining is typically divided into multiple groups, or multiple groups fit with your operational definition. Research suggests there may be a nonlinear relationship between your variables. Past research has established a difference between two groups and adding additional levels to the IV or predictor is warranted.

Some reasons to choose the two-group design:

The variable you are examining is typically divided into two groups (e.g., gender), or two groups fit with your operational definition (e.g., employment status defined as full or part time). You are operationally defining variables in a way that is warranted by the research but has not been tested in the research. The research area you are investigating is relatively new, so that a simple comparison of two groups is warranted. Past research has established the relationship or effect but has been limited to certain populations. You intend to study the relationship or effect in a different population.

A single N study should be chosen when you:

Want to examine a specific cause-effect relationship Have questions about how a manipulation impacts an individual Are using quantitative measures that can be repeated on a daily or weekly basis

A case study should be chosen when you:

Want to gain a holistic sense of a case Have questions about how or why a phenomenon occurred but do not have the ability to control variables Are using primarily qualitative measures that primarily assess past occurrences

3. When would a researcher conduct multiple case studies?

d to build, expand, and generalize theories, s used to determine if results from one single N study generalize to other subjects or participants. to compare patterns across cases

if violate assumptions of parametric stats probability of a ___ error is inc

a Type I error is increased, and we increase the probability that we will reject the null hypothesis when it is true.

5. How is a single N design different from a case study?

a case study is primarily qualitative and is used to understand the whole of a single case, a single N design is a quantitative design used to examine a cause-and-effect relationship within a single case

One way to address the problem of an ABA reversal design is to

add a second manipulation phase (ABAB). Adding additional phases helps to clarify the relationship between the manipulation and the dependent variable.

Nonparametric statistics may be used for a variety of reasons:

as the planned analysis for a study, as a supplement to parametric tests, or because the assumptions of the planned parametric tests could not be met. Nonparametric tests are used with a variety of types of research designs (correlation, two groups, multiple groups, and factorial designs), with the primary reason for their use being the scale of measurement for the dependent variable(s)

1. Why are nonparametric tests less powerful than parametric tests?

because parametric have more stringent requirements less likely to make type 1 error, don't require the typical assumptions about the nature of the underlying distributions that their parametric counterparts do, they are called "distribution free

is often viewed as a purely descriptive or exploratory technique

case study

The most commonly used type of nonparametric test,

chi-square, is used with nominal or categorical data (

can a case generalize to a theory

consensus that a case study can be used to falsify an existing theory. If a theory suggests that something is universal, it takes only one case in which the theory does not apply to falsify that theory. The case can then serve as an important qualifier or encourage modification of the existing theory

The goal of a descriptive study is to

describe characteristics of a population. In order to do so, a representative sample must be obtained from the population. Using probability (random) sampling and having a large sample increases the likelihood that the sample will be representative.

*3. Which of the following types of designs never involves comparing groups within the sample?

descriptive

best choice if your goal is to understand prevalence or trends or to gain in-depth information about a particular phenomenon

descriptive design

1. What are the limitations of sample-based designs?

does not mean it generalizes to everyone

a key advantage of the single N design is the ability to

examine a rare phenomenon that would be impossible to study with a large sample.

The goal of both correlational and experimental studies is to

examine relationships. These designs rely on inferential statistics in order to infer that a relationship found within the sample represents a relationship that exists within the population. A larger sample increases the researcher's ability to find a statistically significant result in the sample when that result exists in the population (i.e., power).

A small N design utilizes a series of single N studies that

examine the same cause-and-effect relationship.

They are the designs of choice when you want to examine causality and it is ethical and feasible for you to manipulate an IV

experiments and quasi

when does the ABA design offer no real advantage over the simple AB design.

if the dependent variable does not revert back to levels comparable to the first baseline after removal of the manipulation,

The single N design also shares a limitation with the case study,

in that its ability to generalize to other individuals or to support or develop a theory is questionable.

When choosing a research design, one important consideration is the balance between

internal validity (i.e., the ability to determine causality) and external validity (i.e., generalizability)

A correlational design examines a __ and useful for

noncausal relationship. It is a useful design when your goal is to predict, but not explain, a phenomenon. A correlational design might also be useful if you are testing a relationship that has not been well researched and you want to establish correlation prior to examining causation, e design of choice when you are testing the reliability or validity of a measurement

when skkwered data are parametic or non para more approp

nonparametric

D. one variable with interval/ratio data and one dichotomous variable

point biserial

The single N design is best grouped with

quasi-experiments because they involve careful observation of the effect of a manipulation without random assignment to condition

A stable baseline has two criteria:

the absence of an upward or downward trend in the data and a small amount of variability

The single N study has some advantages over the case study in this regard because

the assessments are standardized and therefore more easily compared across case

The reversal designs are methodological improvements over the simple AB comparison because ___ but ethical issue

they allow for a clearer understanding of why the behavior changed. However, if improvement occurs in a treatment setting, it may not be ethical to remove the treatment by returning to baseline. Reverting back to the original baseline level can be especially problematic if the original baseline level was severely impacting the individual's quality of life or jeopardizing his or her health.

One of the best reasons to conduct a case study is when your goal is, . A case should be selected because

to gain in-depth knowledge about a particular case or set of cases case study, The goal of a case study is to capture the unique character of the individual case within a real-life context you believe it is prototypical of a certain phenomenon, because it represents an extreme example, or because it is unique and there are few cases available

A baseline must be stable in order for it to be a

useful predictor of future behavior.

can you use spearman in nonparametric

yes Nonparametric tests for ranks compare the sums of the ranks for those in different groups in a study instead of comparing the means (as we would in a parametric test). T