Choosing a Design Pseudoexperimental and True Experimental Designs

Choosing a Design Pseudoexperimental and True Experimental Designs

Research Designs z Experimental designs are concerned with the manipulation of the IV and the measurement of its effect on the DV z Descriptive designs are concerned with the classification of the subjects and with the application of measurement procedures to subjects in order to assess group differences, developmental trends, or relationships among variables.

Research Designs z. Both experimental and descriptive designs are classified as betweensubjects, within-subjects, or mixed designs ythe selection of a design depends largely on the research question

Between-Subject Designs z In experimental designs, different groups of Ss are exposed to different IVs ythe IV is applied to one group (exp) but not to the other group (control) z The major threat to internal validity is subject selection ydifferences may be Ss selection differences rather than the IV y 2 ways to equate experimental and control groups: xrandomization xmatching

Between-Subject Designs z. In descriptive designs, different groups of Ss are compared with each other with regard to their performance on some criterion variable ythe important thing with descriptive designs is to select Ss who fall distinctly into the different categories of the classification variable but who are otherwise equivalent with regard to extraneous variables xclassifications must be constructed that are mutually exclusive

Within-Subjects Designs z. The performance of the same subjects is compared in different conditions. z. In experimental research, the subjects are exposed to all treatment or levels of the IV ybasic concern with these designs is that all conditions should be equivalent except for the application of the IV ytherefore, it is important to control for sequencing or order effect

Within-Subjects Designs z. There are 2 ways to control for the sequencing effect: yrandomization xpresentation of the experimental treatment conditions (IV) is randomly sequenced ycounterbalancing xarrange all possible sequences of treatments (IV) and then randomly assign subjects to each sequence

Within-Subjects Designs z. In descriptive research, longitudinal studies would be a within-subjects design

Mixed Designs z. One IV may be studied with a betweensubjects design while another IV is studied with a within-subjects design ythis is a mixed design

Research Designs z. Two ways to improve generalizability of findings with research designs: yrandom sampling of subjects yreplication

Types of Experimental Designs z. Pseudoexperimental designs (aka “preexperimental”) ydo not have built-in controls ythere may be several explanations for the changes in the DV that are not solely caused by the IV yall of these designs have uncontrolled extraneous variables that threaten the internal validity of the experiment yweaker designs z. True experimental designs

Types of Pseudoexperimental Designs z. One-shot case study z. One-group pretest-posttest z. Static group comparison

One-shot case study z. A descriptive study that has big weakness of no control yno comparisons can be made with this study ystatistics: mainly descriptive statistics (e. g. , means); some inferential statistics (e. g. , correlation coefficients)

One-group pretest-posttest z. Observations are made before and after the IV has been administered to the group ybetter than one shot case study, but still problem with 6 possible uncontrolled extraneous variables (history, maturation, testing, instrumentation, statistical regression, mortality) ystatistics: parametric t-test for correlated samples; nonparametric sign test

Static group comparison z. Comparisons are made between one group which is exposed to the IV and one group which is not yno pretest to compare posttest scores yno way to ensure equivalence between groups on relevant extraneous variables yproblems may arise from subject selection variables, testing, instrumentation, mortality ystatistics: parametric t-test; Mann-Whitney U test; chi-square

True Experimental Designs z. Pretest-Posttest Control Group z. Posttest-Only Control Group z. Solomon Four-Group Design

Pretest-Posttest Control Group z 2 groups of Ss are compared on a measurement or observation on the DV. y. Both groups are measured or observed twice yfirst measurement serves as the pretest and second measurement serves as the posttest yhalf of the Ss are randomly assigned to the first group while the second half are assigned to second group

Pretest-Posttest Control Group z. Only threats to internal validity are testing and mortality zstatistics: independent t-test to compare the two groups; analysis of covariance

Posttest-Only Control Group z. Identical to the pretest-posttest control group design except that the pretest is not administered to either of the two groups yby random assignment of Ss to the two groups, it controls selection, history, maturation, and statistical regression ytesting and instrumentation do no exist since none of the Ss is measured twice ymortality could be an internal validity problem

Solomon Four-Group Design z. Subjects are randomly assigned to four different groups ytwo of the groups receive the treatment (IV) ytwo of the groups do not receive the treatment yonly one of the control groups is administered the pre-test yall four groups are administered the posttest

Solomon Four-Group Design z. It is a combination of the pretest-posttest and posttest-only control group designs zit controls for all threats to internal validity zstatistics: 2 -way ANOVA

What’s Coming Up …. z Human Subject Certification test (next week) z Group Work (next week) z Specific Group Designs (2 weeks) z Single-subject Designs (3 weeks) z Intro/Methods (Oct. 28) z JOURNAL CLUB (Nov 4) z Statistics y descriptive y inferential z Miscellaneous y interpreting research y role of research in CDIS

SLP Journal Club z 9 groups y 8 groups with 3 members y 1 group with 2 members z one group presentation/week z recommendation: select article based on your research project

AUD Journal Club z 2 groups y 2 groups with 2 members z one group presentation/week z recommend select article from research project