Regression analysis with observational data remains the primary analytical method in political science.
However, we will never know whether we controlled for all unobserved confounders.
New sensitivity tools help to assess how likely results are robust to missing controls.
However, we will never know whether we controlled for all unobserved confounders.
New sensitivity tools help to assess how likely results are robust to missing controls.
The robustness value (@analisereal,@chadhazlett) applies to multivar. regressions: It is the minimum strength of association that unobserved confounding would need to have, both with the treatment and with the outcome, to change the research conclusions.
The robustness value can be calculated with a Shiny Application (carloscinelli.shinyapps.io/robustness_val…) or with the #rstats package sensemakr.
Second, the E-value applies to risk ratios: It is the minimum strength of association on the risk-ratio scale that an unmeasured confounder would need to have with both the treatment assignment and the outcome to fully explain away a specific treatment-outcome association [...]
A new #Stata ado to calculate E-values by (@_MMathur et al.) was just released journals.sagepub.com/doi/full/10.11…
Both approaches help us to better communicate not only what we know but also how certain or uncertain we can be about our knowledge claims.
Addendum: There is also an R package on CRAN to compute E-values cran.r-project.org/web/packages/E…
Addendum 2: And they also have a Shiny Web Application!
Both for single studies (
evalue-calculator.com ) and meta-analyses ( mmathur.shinyapps.io/meta_gui_2/ )
Both for single studies (
evalue-calculator.com ) and meta-analyses ( mmathur.shinyapps.io/meta_gui_2/ )
