User Tools

Site Tools


development:best_practices_estimation

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
Next revision
Previous revision
development:best_practices_estimation [2016/04/20 09:17]
schuemie
development:best_practices_estimation [2020/03/09 05:18]
schuemie
Line 7: Line 7:
  
   * **Transparency**:​ others should be able to reproduce your study in every detail using the information you provide.   * **Transparency**:​ others should be able to reproduce your study in every detail using the information you provide.
 +
   * **Prespecify** what you're going to estimate and how: this will avoid hidden multiple testing (fishing expeditions,​ p-value hacking). Run your analysis only once.   * **Prespecify** what you're going to estimate and how: this will avoid hidden multiple testing (fishing expeditions,​ p-value hacking). Run your analysis only once.
 +
   * **Validation of your analysis**: you should have evidence that your analysis does what you say it does (showing that statistics that are produced have nominal operating characteristics (e.g. p-value calibration),​ showing that specific important assumptions are met (e.g. covariate balance), using unit tests to validate pieces of code, etc.)   * **Validation of your analysis**: you should have evidence that your analysis does what you say it does (showing that statistics that are produced have nominal operating characteristics (e.g. p-value calibration),​ showing that specific important assumptions are met (e.g. covariate balance), using unit tests to validate pieces of code, etc.)
  
 ===== Best practices (generic) ===== ===== Best practices (generic) =====
- +  * **Write a full protocol**, and make it public prior to running the study. This should include 
- +    * Research question + hypotheses to be tested 
-  Make all analysis code available as **open source*+    Which method(s), data, cohort definitions.  
 +    ​What is the primary analyses and what are sensitivity analyses? 
 +    ​Quality control 
 +    ​Amendments and Updates
  
   * **Validate** all code used to produce estimates. The purpose of validation is to ensure the code is doing what we require it to do. Possible options are:   * **Validate** all code used to produce estimates. The purpose of validation is to ensure the code is doing what we require it to do. Possible options are:
Line 25: Line 30:
   * Produce **calibrated p-values**   * Produce **calibrated p-values**
  
 +  * Make all analysis code available as **open source** so others can easily replicate your study
  
 ===== Best practices (new-user cohort design) ===== ===== Best practices (new-user cohort design) =====
  
  
 +  * Use **propensity scores** (PS)
  
 +  * Build PS model using **regularized regression** and a **large set of candidate covariates** (as implemented in the CohortMethod package)
  
 +  * Use either **variable-ratio matching** or **stratification** on the PS
 +
 +  * **Compute covariate balance** after matching for all covariates, and terminate study if a covariate has standardized difference > 0.1
  
  
 ===== Best practices (self-controlled case series) ===== ===== Best practices (self-controlled case series) =====
- 
  
   * Include a **risk window just prior to start of exposure** to detect time-varying confounding (e.g. contra-indications,​ protopathic bias)   * Include a **risk window just prior to start of exposure** to detect time-varying confounding (e.g. contra-indications,​ protopathic bias)
 +
 +===== Best practices ((nested) case-control) =====
 +
 +  * **Don'​t** do a case-control study
development/best_practices_estimation.txt · Last modified: 2020/03/09 05:18 by schuemie