Lecture+5-+quasi+experimental+design

Lecture+5-+quasi+experimental+design - SE 10 Research...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: SE 10 Research Design Lecture 5 Quasi Experimental Design 1 Class project info Assignment is posted on the class website You will be evaluating a published article based on the topics we've mentioned in class Pretty short (56 pgs) Assignment will have lots of structure you can write some sections before we cover others Variable definitions, measurement, internal validity, etc. 2 Quasiexperimental design Similar to true experiments except it does not use random assignment Designs with only one group are okay as long as there is a pre and post test Often a lot more practical in the social sciences 3 One group pre/post test design O1 X1 O2 Allows for comparison only within subjects May be only practical design if withholding treatment is unethical 4 Nonequivalent groups post test design X1 O1 O1 dashed line means no random assignment Allows for comparison only between subjects 5 Pre/Post test design O1 O1 X1 O2 O2 Pre test allows you to test for group similarity on characteristics you think are important Between subjects and within subjects comparison 6 Interrupted time series O1 O2 O3 O4 X1 O5 O6 O7 Additional measurements help ensure that you aren't just capturing a phenomenon that is occurring regardless of your X The "interruption" just means X Allows for comparisons within subjects 7 Multigroup interrupted time series design O1 O1 O2 O2 O3 O4 X1 O3 O4 O5 O6 O5 O6 O7 O7 Allows for comparison between groups and within groups 8 How do we assign groups in quasi experimental designs? Short answer: it depends Methods exist for increasing the equivalency of the control and experimental group Is X something you can manipulate or not? Can you do a pretest? Matched pairs Pretests 9 If you can manipulate X... The independent variable (x) is not a pre existing condition, but something you will apply If pretests are possible If pretests are not possible Measure subjects on variables you think matter and split like subjects up equally Select groups that are likely to be similar on important characteristics 10 If X cannot be manipulated... The independent variable may be a preexisting condition or a program already in place If pretests are possible If pretests are not possible Measure subjects on important characteristics and hand select a control group who is similar except for the independent variable Select groups likely to be similar Problem is you know they are different a good control group might not be available 11 Assignment techniques: matched pairs Select multiple pairs of subjects who have very similar characteristics From each pair, assign one to control group and one to experimental group Similar paired subjects may also be similar on unmeasured or unobserved characteristics 12 Assignment techniques: pre tests Measure all subjects on important characteristics Hand select groups so characteristics are balanced Does not account for unmeasured or unobserved characteristics 13 Midterm review important concepts Conceptualizing, operationalizing Variables Measurement reliability and validity Internal validity, threats to validity Alternative hypotheses Experimental Design 14 ...
View Full Document

Ask a homework question - tutors are online