Arming Oneself With NOIR (Don't you mean NAIR, Jim?)
Data Type Identification Primer, Terminology Refresher
Statistical Primer Information
The following information is designed to enhance your decision-making as you work through the research scenarios and flow charts. For additional elaboration about particular concepts, please refer to the glossary (in text).
Nominal Data---variables are classified into mutually exclusive and exhaustive categories. While researchers may assign some number to these variables for computerized analysis (e.g., support group participation = 1, no support group participation = 2), nominal variables have no inherent numerical value. Thus, we could have used 1,000 and 2,000 respectively in the preceding example. This type of data also presents the fewest analytical options compared to the remaining data types.
Ordinal Data---variables are classified into qualitatively different categories (somewhat similar to their "cousin" nominal) and the values of the variables represent a ranking. For example, an investigator could ask several department managers to identify their five best employees; Joe = first; Sara = second; Chase = third, etc. Of course, the managers could also reply that Joe is outstanding; Sara is excellent; and Chase is promising. While researchers can assign numbers to represent the rankings (e.g., outstanding = 1; excellent = 2; and promising = 3), this information is limited as we cannot truly determine the difference between the ranks. Hence, we could have used 500, 490, and 480 respectively in the preceding example. The focus here, then, is on the actual ranking and the number of times it occurs. This type of data also presents more analytical options than the nominal type, but fewer analytical options than interval or ratio types.
Interval Data---variables are classified into qualitatively different categories, ranked along some dimension, and are established as having standard or equal distances between each of the adjacent categories/levels. The most common example of this type of data is found in Likert scales commonly used in survey research. Individuals read and respond to a series of statements according to the following choices: "strongly agree (SA)," "agree," "undecided/neutral," "disagree," or "strongly disagree (SD)." In this five-point example, the numbers, 1-5, would represent the respective choices in ascending or descending order. For instance, if SA = 1, then SD = 5; conversely, if SA = 5, then SD = 1. Remember, unlike nominal and ordinal data types, the presumption here is that these distances are meaningful from an analytical vantage point. This type of data also presents more analytical options than the nominal and ordinal types.
Ratio Data---variables are classified into qualitatively different categories, ranked along some dimension, are established as having standard or equal distances between each of the adjacent categories/levels, and an absolute zero point indicates where the variables cease to exist. An example is the amount of time spent watching prime-time TV dramas on a daily basis. The variable, time, could range from 0 to some amount of hours such as 3. Another example could be the amount of eye contact (ECA) between instructors and students during exams. ECA could range from 0 to some amount of minutes or hours. Similar to interval data, the presumption here is that these distances are meaningful from an analytical vantage point. This type of data also presents more analytical options than the nominal and ordinal types.
Testing for differences or associations---The tests for difference usually take the form of hypotheses (see below). The goal in hypothesis testing is to provide empirical evidence (not the same as absolute truth!) that the examined variables vary among the study's participants. The tests for association usually take the form of research questions (RQs). The goal in RQ testing is to establish that a relationship exists between the examined variables among the study's participants.
A Closer Look at RQs and Hypotheses (Hys)
An RQ advances a suspected relationship between two variables by posing a research problematic. RQS usually occur in the early stages of a project, or when there is a paucity of research. Some examples include: Does a relationship exist between stress and disconfirming communication among nurses? Is there an association between the HMO's climate and client satisfaction? What are the recurring themes of well-adjusted PWAs within a community living facility? The researchers are not sure that any relationship between the variables exists. Due to this uncertainty, the alpha level of RQs is often set at .10.
Brief Review About HYs
A non-directional hy provides a two-tailed test of a suspected relationship. Although stronger than an RQ, it is less powerful than a directional hy. An example would be: A cohesive HMO climate is related to client satisfaction. A non-directional hy thus encompasses both possibilities, a positive and negative relationship. Its alpha level is most often .05/2.
In contrast, a directional hy provides a one-tailed test and might look like this: A highly cohesive HMO climate produces high client satisfaction. It is the most powerful test of the variables and its alpha level is .05. Why is this test the most powerful?? Let's consider a roulette table example. With a non-directional hypothesis, the researcher is betting on two outcomes (positive and negative). With a directional hy, he/she is only betting on only one outcome (either positive or negative).
The final type of hypothesis is called the null hypothesis. This declarative statement states that there is no difference nor association among the examined variables within a particular study. While many contemporary studies no longer state the null, it is always being tested. Once researchers have analyzed their data and calculated the appropriate test statistic, they then make a decision about the null hypothesis. This decision has only two possibilities: rejection of the null or embracing (accepting) the null. Proceeding from this decision, researchers then make a decision about their RQ or hypothesis.
Terminology Refresher
Control Group---a collection of individuals who do not receive the manipulation of independent variable(s). They may receive, however, a placebo.
Dependent variable---a target (e.g., behavior, event, process) which measures the impact of manipulating the independent variable or the extent of the independent variable's influence.
Experiment---an investigation which involves independent variables, pre-and-post tests and a control group.
Independent variable---a target (e.g., behavior, event, process) which is manipulated by the investigator(s) or presumed to influence some outcome measure.
Kurtosis---deviance in the amplitude of a distribution from a normal curve. If the curve is narrow and sharply pointed, it is said to be peaked. If the scores do not cluster around the middle, but are dispersed rather evenly across the distribution, the curve is said to be flat.
Pre-test---a data collection measure administered prior to an investigation.
Post-test---a data collection measure administered after an investigation.
Probability level---the level at which hypotheses and/or RQs are tested (e.g., .05, .025, .10).
Skewness---the asymmetry that occurs when the majority of scores fall toward one end of a distribution, making the curve tail off in that direction. Positive skew occurs when the tail runs to the right side of the curve; negative skew occurs when the tail runs to the left side of the curve.
Variable---any research concept that takes two or more values; the target item(s) of a study.
The following information is designed to enhance your decision-making as you work through the research scenarios and flow charts. For additional elaboration about particular concepts, please refer to the glossary (in text).
Nominal Data---variables are classified into mutually exclusive and exhaustive categories. While researchers may assign some number to these variables for computerized analysis (e.g., support group participation = 1, no support group participation = 2), nominal variables have no inherent numerical value. Thus, we could have used 1,000 and 2,000 respectively in the preceding example. This type of data also presents the fewest analytical options compared to the remaining data types.
Ordinal Data---variables are classified into qualitatively different categories (somewhat similar to their "cousin" nominal) and the values of the variables represent a ranking. For example, an investigator could ask several department managers to identify their five best employees; Joe = first; Sara = second; Chase = third, etc. Of course, the managers could also reply that Joe is outstanding; Sara is excellent; and Chase is promising. While researchers can assign numbers to represent the rankings (e.g., outstanding = 1; excellent = 2; and promising = 3), this information is limited as we cannot truly determine the difference between the ranks. Hence, we could have used 500, 490, and 480 respectively in the preceding example. The focus here, then, is on the actual ranking and the number of times it occurs. This type of data also presents more analytical options than the nominal type, but fewer analytical options than interval or ratio types.
Interval Data---variables are classified into qualitatively different categories, ranked along some dimension, and are established as having standard or equal distances between each of the adjacent categories/levels. The most common example of this type of data is found in Likert scales commonly used in survey research. Individuals read and respond to a series of statements according to the following choices: "strongly agree (SA)," "agree," "undecided/neutral," "disagree," or "strongly disagree (SD)." In this five-point example, the numbers, 1-5, would represent the respective choices in ascending or descending order. For instance, if SA = 1, then SD = 5; conversely, if SA = 5, then SD = 1. Remember, unlike nominal and ordinal data types, the presumption here is that these distances are meaningful from an analytical vantage point. This type of data also presents more analytical options than the nominal and ordinal types.
Ratio Data---variables are classified into qualitatively different categories, ranked along some dimension, are established as having standard or equal distances between each of the adjacent categories/levels, and an absolute zero point indicates where the variables cease to exist. An example is the amount of time spent watching prime-time TV dramas on a daily basis. The variable, time, could range from 0 to some amount of hours such as 3. Another example could be the amount of eye contact (ECA) between instructors and students during exams. ECA could range from 0 to some amount of minutes or hours. Similar to interval data, the presumption here is that these distances are meaningful from an analytical vantage point. This type of data also presents more analytical options than the nominal and ordinal types.
Testing for differences or associations---The tests for difference usually take the form of hypotheses (see below). The goal in hypothesis testing is to provide empirical evidence (not the same as absolute truth!) that the examined variables vary among the study's participants. The tests for association usually take the form of research questions (RQs). The goal in RQ testing is to establish that a relationship exists between the examined variables among the study's participants.
A Closer Look at RQs and Hypotheses (Hys)
An RQ advances a suspected relationship between two variables by posing a research problematic. RQS usually occur in the early stages of a project, or when there is a paucity of research. Some examples include: Does a relationship exist between stress and disconfirming communication among nurses? Is there an association between the HMO's climate and client satisfaction? What are the recurring themes of well-adjusted PWAs within a community living facility? The researchers are not sure that any relationship between the variables exists. Due to this uncertainty, the alpha level of RQs is often set at .10.
Brief Review About HYs
A non-directional hy provides a two-tailed test of a suspected relationship. Although stronger than an RQ, it is less powerful than a directional hy. An example would be: A cohesive HMO climate is related to client satisfaction. A non-directional hy thus encompasses both possibilities, a positive and negative relationship. Its alpha level is most often .05/2.
In contrast, a directional hy provides a one-tailed test and might look like this: A highly cohesive HMO climate produces high client satisfaction. It is the most powerful test of the variables and its alpha level is .05. Why is this test the most powerful?? Let's consider a roulette table example. With a non-directional hypothesis, the researcher is betting on two outcomes (positive and negative). With a directional hy, he/she is only betting on only one outcome (either positive or negative).
The final type of hypothesis is called the null hypothesis. This declarative statement states that there is no difference nor association among the examined variables within a particular study. While many contemporary studies no longer state the null, it is always being tested. Once researchers have analyzed their data and calculated the appropriate test statistic, they then make a decision about the null hypothesis. This decision has only two possibilities: rejection of the null or embracing (accepting) the null. Proceeding from this decision, researchers then make a decision about their RQ or hypothesis.
Terminology Refresher
Control Group---a collection of individuals who do not receive the manipulation of independent variable(s). They may receive, however, a placebo.
Dependent variable---a target (e.g., behavior, event, process) which measures the impact of manipulating the independent variable or the extent of the independent variable's influence.
Experiment---an investigation which involves independent variables, pre-and-post tests and a control group.
Independent variable---a target (e.g., behavior, event, process) which is manipulated by the investigator(s) or presumed to influence some outcome measure.
Kurtosis---deviance in the amplitude of a distribution from a normal curve. If the curve is narrow and sharply pointed, it is said to be peaked. If the scores do not cluster around the middle, but are dispersed rather evenly across the distribution, the curve is said to be flat.
Pre-test---a data collection measure administered prior to an investigation.
Post-test---a data collection measure administered after an investigation.
Probability level---the level at which hypotheses and/or RQs are tested (e.g., .05, .025, .10).
Skewness---the asymmetry that occurs when the majority of scores fall toward one end of a distribution, making the curve tail off in that direction. Positive skew occurs when the tail runs to the right side of the curve; negative skew occurs when the tail runs to the left side of the curve.
Variable---any research concept that takes two or more values; the target item(s) of a study.
Nikki's Veranda
![[Create your own website at www.homestead.com!]](../../defaultUser/images/footer.gif)