Lecture 08 - PAM 3300: Research Designs and Causal Claims...

Info iconThis preview shows pages 1–7. Sign up to view the full content.

View Full Document Right Arrow Icon
PAM 3300: Research Designs and Causal Claims
Background image of page 1

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

View Full DocumentRight Arrow Icon
Research Design For our purposes, research design refers to the set of analyses that will be implemented to assess the causal effect of X on Y. The design measured. Most importantly, the research design clarifies how the outcomes of observations with different levels of X will be compared to one another to relate X to Y. The design used is determined by: How data was collected - in other words, the reasons for variation in X in the data. Availability of data on control groups and availability of data from pre- and post-treatment outcomes. The availability of other important information about observations.
Background image of page 2
Research Design: some options 1. Randomized control trials Units are randomly chosen to be given the treatment/program/drug, etc. The gold standard, but not perfect. 2. Post Treatment w/Comparison Group Compare post-treatment outcomes for people who did and did not receive the treatment. 3. Pre/Post Treatment (reflexive controls) Compare outcomes of a a group before and after receiving a treatment. 4. Pre/Post Treatment with Comparison Group Compare outcomes of two groups before and after treatment for a group that does and a group that does not receive the treatment. 5. Interrupted time-series. Similar to pre-post design, only with many years of ‘pre-’ and many years of ‘post-’ Note: All of these designs may or may not use regression to control for other variables.
Background image of page 3

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

View Full DocumentRight Arrow Icon
Research Design: more options 6. Matching Outcomes for people in the treatment group are compared to someone in the control group who looks most like a match in terms of their characteristics. 7. Natural Experiments: difference-in-differences Utilize a change in policy or some natural event that approximate random assignment of a treatment. 8. Instrumental Variables Find another variable that is uncorrelated with things in the error term, but is correlated with whether one receives the treatment. Analyze the impact of changes in that ‘instrumental variable’ to get at the true effect of the treatment on the outcome. 9. Regression Discontinuity The gold standard of non-experimental methods.
Background image of page 4
Validity The research design determines the “validity” of causal claims. Validity is simply the extent to which we believe a claim - a measure of its “truthiness.” Kinds of validity: Construct validity Statistical conclusion validity Internal validity External validity
Background image of page 5

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

View Full DocumentRight Arrow Icon
Do the variables in the data set adequately capture the concepts/constructs we are trying to learn about? Ex: Does having more ‘social capital’ improve health? Examine by regressing county-level disease prevalence rates on
Background image of page 6
Image of page 7
This is the end of the preview. Sign up to access the rest of the document.

Page1 / 28

Lecture 08 - PAM 3300: Research Designs and Causal Claims...

This preview shows document pages 1 - 7. Sign up to view the full document.

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