Thursday, December 11, 2014

Experimental Research Design & Threats to Validity


The Experimental Research Design

Overview

There are 3 types of research designs: experimental, non-experimental, and quasi-experimental. Despite the type of research design, experiments involve first taking action and then observing the consequences of that action.

What is an Experimental Design?

The experimental research design is the gold standard as it is the only type of research design that meets all 3 criteria for causality (that is, empirical association, temporal order, non-spuriousness). It is the best design to account for causality. 

The experimental design determines the cause of something by isolating the hypothesized cause and comparing the controlled results. It has high internal validity; may not be strong with external validity as this type of design is highly controlled so much that it may not reflect reality. Construct validity may also be a concern.

3 Components of an Experimental Design

There are 3 components of an experimental design, as follows:

   1.) manipulation of the independent variable (also called the treatment variable or the possible cause)

   2.) pre-test and post-test of the dependent variable (the possible effect)

   3.) randomization aka random assignment into the experimental/treatment group and control group; the most important characteristic of an experimental research design

The Experimental Group versus The Control Group

- experimental group: the group exposed to the independent variable/those whom receive the treatment

- control group: exposed to all the same of factors of experimental group but does not receive the independent variable / treatment

Randomization (Key to the Experimental Design!)

Because of randomization, both the experimental and control group are equal; their sample statistics are the same or in other words, the experimental and control group are statistically equivalent and they look identical; with randomization, you are controlling for all other factors (non spuriousness); also because of randomization, the possible independent variables become constants; random assignment makes it possible to assume that all other things are equal among the two groups

Possible problems with experimental research designs:

ethics - giving one group the treatment and the other not may pose a problem
time and expense - experimental designs more expensive than other designs
selection bias - 
placebo effect - 
experimenter bias - experimenters may have a disposition to see results a certain way; a double blind experiment can control for this

Threats to internal validity:

history - historical events may occur during the experiment that confound the results of the study; this represents something that occurs outside the study

maturation - people mature and this may change results; refers to change within the subject themselves; "the changes to a respondent during the course of a study that can influence the findings" (Lanier & Briggs, 2011)

testing effect - people knowing they are being tested may alter their behavior/responses; "typically occurs in presetting and possessing studies when the familiarity of the test instrument influences the outcome of the posttest" (Lanier & Briggs, 2011)

instrumentation- concerned with changes in the measurement process itself; how we measure something; :a threat to the quality of a study when changes are made to to the testing instrument" (Lanier & Briggs, 2011)

statistical regression (also called regression to the mean)- can occur when researchers are interested in cases that have extreme scores on some variable; "when statistical outliers move closer to the average response" (Lanier & Briggs, 2011)

selection biases- for example, if volunteers are studied, the volunteers may be very different from other types of people; "occurs in the sampling process when selection to a research group or assignment to research groups, experimental or control, is poorly done and compatibility is lacking" (Lanier & Briggs, 2011)

experimental mortality (also called attrition)- occurs when people drop from the study, die, or otherwise become 'missing'

time order- when there is ambiguity about the time order of the experimental stimulus and the dependent variable

diffusion or imitation of treatments- when the subjects of the experimental group and control group are able to communicate with each other, it is possible that the experimental subjects may pass on some elements of the experimental stimulus (the independent variable) to the control group, causing spillover effects

compensatory treatment- may occur when participants (such as police officers or probation officers) give special treatment to the control group to make up for treatment they aren't receiving as opposed to the experimental group

compensatory rivalry- subjects who do not receive the experimental stimulus may try to compensate by working harder; based on the subjects' reactions to the experiment

demoralization- subjects in the control group may give up as they may feel depraved

experimental mortality - also called attrition


--------------------------------------------------------------------------------------------------
                                      Pretest*                                       Posttest
                                      Time 1             Time 2              Time 3

Group 1
Experimental Group     DV                     IV                     DV

Group 2                      
Control Group               DV                                              DV


* The pretests should be about the same
--------------------------------------------------------------------------------------------------

Different types of experimental designs

   1.) classical design - subjects are divided into at least 2 groups - experimental group and control group; minimum of 2 time frames

   2.) after-only design - here, the classical design is cut in half by removing the initial time frame; since the initial time frame is removed, we can no longer be sure that the experimental group and the control group are indeed statistically significant prior to exposing the experimental group to the independent variable


Research Methods in Criminal Justice and Criminology A Mixed Methods Approach
Mark M. Lanier and Lisa T. Briggs
2011


No comments:

Post a Comment