Causal Inference and Robustness Indices
Presentations
Quick examples [pdf of quick examples] 30 Minutes
powerpoint for combined frameworks 3 hours (pdf)
powerpoint for replacement of cases [pdf of replacement of cases] 2 hours
full publishable write-up for case replacement
powerpoint for correlation framework [pdf of correlation framework] 1 hour
full publishable write-up for correlation
powerpoint for Case Replacement for Logistic Regression 10 minutes
powerpoint for ordered thresholds relative to transaction costs 10 minutes
powerpoint for alternative scenarios and related techniques 15 minutes
powerpoint for comparison of frameworks [pdf of comparison of frameworks]
Software
R
Excel Spreadsheet
spreadsheet for calculating indices [KonFound-it!©]
STATA
Example data sets
Anscombe’s Quartet.dta (courtesy of CHATTERJEE, HADI, & PRICE)
Anscombe’s Quartet.rds (courtesy of CHATTERJEE, HADI, & PRICE)
You can also copy into c:\ado\personal (created by Ran Xu)
PKonFound (sensitivity for published results)
KonFound (sensitivity for a user run model in Stata)
Requires:
. ssc install indeplist
. ssc install matsort
MKonFound (meta sensitivity across multiple studies)
SAS
Articles
Frank, K.A., Maroulis, S., Duong, M., and Kelcey, B. 2013. What would it take to Change an Inference?: Using Rubin’s Causal Model to Interpret the Robustness of Causal Inferences. Education, Evaluation and Policy Analysis. Vol 35: 437-460.
When one reports a regression from an observational study, there is always the concern: "But have you controlled for xxx?" In this article, I index the impact of a confounding variable in terms of the product of two correlations (correlation between the confounding variable and the predictor of interest X correlation between the confounding variable and the outcome -- see figure below). This product appears in both the estimate of a regression coefficient and its standard error. Thus it can be used to quantify the impact necessary to alter a statistical inference. In the article I develop the index for the bivariate and multivariate contexts, compare it to other statistics, apply to an example, describe a reference distribution, and then interpret in the context of recent concerns about causal and statistical inferences.
see also
Frisco, Michelle, Muller, C. and Frank, K.A. 2007. Using propensity scores to study changing family structure and academic achievement. Journal of Marriage and Family. Vol 69(3): 721–741
Concerns about Causality in the Network Influence Model (ppt)