Lecture 11

Lecture 11 - PAM 3300 Non-Experimental Designs Designs with...

Info iconThis preview shows pages 1–8. Sign up to view the full content.

View Full Document Right Arrow Icon
PAM 3300: Non-Experimental Designs Designs with no control group or no pretest
Background image of page 1

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
What are the threats to validity? One explanation for the findings is that students this year are ‘less able’ than the previous cohort of students, and this caused the decline in test scores. But there are other plausible alternative explanations for findings. These need to be ruled out in order to make the claim more credible. What are other possible explanations? (selection) changing composition of test-takers: State requirements that all children take the exam Average scores for those planning on applying for financial aid: 501 reading-508 math; 530/548 for those who don’t plan to apply. (history) different testing instrument (new writing section in 2006) (selection) fewer students take the test more than once (gain is usually 14pts/section for 2nd try; 10 pts/section for 3rd try). others
Background image of page 2
How might you rule out alternative explanations? What data would you need? Some alternative explanations might be ruled out by looking at patterns of outcomes. Test scores for middle income whites Math scores may be unaffected by switch in format of writing exam. Limit the analysis to first time test-takers.
Background image of page 3

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is an example of a pre-test post-test with no comparison group design Or, in “design notation” In design notation: each group gets their own row Observations on outcomes are denoted by ‘O’ (subscripts for different measures, or time periods, or groups) Treatments are denoted ‘X’ Other notation: ‘N’ or ‘R’ to indicate whether groups are randomly assigned when there is more than one group O 1 X O 2
Background image of page 4
Another example: The NYC Voucher experiment In “design notation” In design notation: Groups are randomly assigned, and given pre-test. First group is offered a voucher to attend private school, second is not. Post-test measures are collected for 3 years. O A0 X O A1 O B0 O B1 R R O A2 O B2 O A3 O B3
Background image of page 5

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Research designs with no pre-test One of the most popular research designs involves correlating some variables X and Y in a cross-sectional dataset. Examples abound: Washing after sex increases the risk of HIV (data on different men in Africa; washers contract HIV at higher rate than non- washers) Red wine reduces the risk of heart attack (data on adults in U.S.; red wine drinkers have fewer heart attacks than non-red wine drinkers) More generous welfare programs reduce work effort among single moms (data on states; states with higher welfare payments have higher unemployment rates for single mothers)
Background image of page 6
In “design notation” Groups are not randomly assigned. Measures of outcomes only in for one year following the treatment.
Background image of page 7

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
Image of page 8
This is the end of the preview. Sign up to access the rest of the document.

This note was uploaded on 03/28/2009 for the course PAM 3300 taught by Professor Matsudaira during the Spring '08 term at Cornell.

Page1 / 26

Lecture 11 - PAM 3300 Non-Experimental Designs Designs with...

This preview shows document pages 1 - 8. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online