GCU Core Quantitative Designs

Design

Description

General Requirements

Non-Experimental Class

Goal is to examine relationships between variables or comparisons among groups, with no treatment or intervention involved. Causal conclusions cannot be reached using non-experimental designs.

Correlational or Associative

Examines relationship(s) between pairs of variables using data from a single group of participants with the intent of assessing the direction and strength of a relationship. Can use primary (i.e., collected by the learner) and/or secondary data (i.e., not collected by the learner).

· Examines relationships between variables in a naturally occurring setting. Variables should not or cannot be manipulated.

· There is a theoretical and/or research-based justification for expecting a correlation or association.

· Requires valid approaches to data collection such as validated surveys or databases.

· Can use categorical (ordinal or nominal) or continuous (interval or ratio) variables.

· Data analysis involves some type of correlation or association test.

Correlational-predictive

Examines relationship(s) between two or more variables using data from a single group of participants, with the intent of predicting a criterion variable from one or more predictor variables. Can use primary (i.e., collected by the learner) and/or secondary data (i.e., not collected by the learner).

· Examines relationships between variables in a naturally occurring setting. Variables should not or cannot be manipulated.

· There is a theoretical and/or research-based justification for expecting a predictive relationship.

· Requires valid approaches to data collection such as validated surveys or databases.

· Can use categorical (ordinal or nominal) or continuous (interval or ratio) variables.

· Data analysis will require some type of regression.

Comparative

Examines differences between two or more groups defined by one or more categorical variables and/or between two or more measurements of a single group. Uses primary data (i.e., collected by the learner) and there is no manipulation of variables.

· Examines relationships between variables in a naturally occurring setting. Variables should not or cannot be manipulated.

· There is a theoretical and/or research-based justification for expecting differences.

· Requires valid approaches to data collection such as validated surveys.

· Define (choose for comparison) mutually exclusive groups that are as homogeneous as possible.

Ex Post Facto

Examines differences between two or more groups defined by one or more categorical variables and/or between two or more measurements of a single group. Uses secondary data (i.e., not collected by the learner).

· Examines relationships between variables in a naturally occurring setting. Variables should not or cannot be manipulated.

· There is a theoretical and/or research-based justification for expecting differences.

· Requires data justifiably assumed to be valid and reliable.

Experimental Class

Goal is to examine the effect(iveness) of some treatment / intervention.

True

Experimental

Examines the effect/outcome of some form of treatment(s) using random assignment of participants to treatment and control groups. The researcher controls both treatment and measurement. Uses primary data.

· Two or more equivalent groups to receive one or more clearly defined and controlled treatment(s) and a control group.

· Random assignment of participants to each of the groups.

· Standardization of all aspects of research procedures employed to ensure conditions are the same for all participants (e.g., control of potential confounding variables).

· Categorical independent variable(s) and interval or ratio level dependent variable(s).

· Strongest in terms of both internal and external validity.

· Strong support for cause-and-effect conclusions.

Pre-Experimental

Examines the effect/outcome of some form of treatment(s) using either one or two pre-existing group of participants. May use a one-shot comparison group (not a control group measured pre and post). Uses primary data (i.e., data collected by the learner).

· Only two such designs generally will be considered:

· One Group Pretest-Posttest Design

· O X O

· A single group is measured both before and after some treatment. No comparison to a non-treatment group is made.

· Static Group Comparison Design

· O X O

O

· A group is measured after some treatment, and this result is compared to a measurement from a second group that did not receive the treatment.

· Typically, no random assignment – participants are in pre-existing groups or groups that are naturally formed (group inclusion is beyond the control of the researcher).

· Conducted with similar rigor and control as experimental studies with clearly defined treatments.

· Requires categorical independent variable(s) and ordinal, interval or ratio level dependent variable(s).

· Potential for many confounding variables.

· Extremely “weak” in terms of both internal and external validity.

· Very little support for cause-and-effect conclusions.

· Often used for exploratory research.

· Should be considered only if other experimental designs are not an option.

Quasi-Experimental

Examines the effect/outcome of some form of treatment(s) using either one or two pre-existing group of participants. May use a control group measured pre and post. Uses primary data (i.e., data collected by the learner.

· The researcher selects the groups to compare and when to make the measurements.

· Some example designs:

· Non-Equivalent Control Groups Design

· O X O

· O O

· Simple Interrupted Time Series Design

· O O O O X O O O O

· Typically, no random assignment – participants are in pre-existing groups or groups that are naturally formed (group inclusion is beyond the control of the researcher).

· Conducted with similar rigor and control as experimental studies with clearly defined treatments.

· Requires categorical independent variable(s) and ordinal, interval or ratio level dependent variable(s).

· Potential for many confounding variables.

· Better than pre-experiment in terms of both internal and external validity.

· Somewhat stronger support for cause-and-effect conclusions.

· Should be considered only if a true experiment is not an option.