Differences

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

--- projects:workgroups:patient-level_prediction:best-practice [2016/05/03 18:25] – [General principles] prijnbeek
+++ projects:workgroups:patient-level_prediction:best-practice [2016/05/04 15:43] (current) – [Best practices] prijnbeek
@@ Line 6: / Line 6: @@
-  * **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. Make sure all analysis code is available as open source.
-  * **Prespecify** what you're going to predict and how: this will avoid fishing expeditions, p-value hacking. Run your analysis only once against the test set.
+  * **Prespecify** what you're going to predict and how. This will avoid fishing expeditions, p-value hacking.
-  * **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.)
+  * **Code validation**: it is important to add unit tests, code review, or double coding steps to validate the developed code base. We recommend to test the code on benchmark datasets.
+===== Best practices =====
-===== Best practices (generic) =====
+**Data characterisation and cleaning**: Before modelling it is important to characterize the cohorts, for example by looking at the prevalence of certain covariates. Tools are being developed in the community to facilitate this. A data cleaning step is recommended, e.g. remove outliers in lab values.
+**Dealing with missing values **: A best practice still needs to established.
+**Feature construction and selection**: Both feature construction and selection should be completely transparent using a standardised approach to be able repeat the modelling but also to enable application of the model on unseen data.
+**Inclusion and exclusion criteria** should be made explicit. It is recommended to do sensitivity analyses on the choices made. Visualisation tools could help and this will be further explored in the WG.
+**Model development** is done using a split-sample approach. The percentage used for training could depend on the number of cases, but as a rule of thumb 80/20 split is recommended. Hyper-parameter training should only be done on the training set.
+**Internal validation** is done only once on the holdout set. The following performance measures should be calculated:
+  . Overall performance: Brier score (unscaled/scaled)
+  . Discrimination: Area under the ROC curve (AUC)
+  . Calibration: Intercept + Gradient of the line fit on the observed vs predicted probabilities
+We recommend box plots of the predicted probabilities for the outcome vs non-outcome people, the ROC plot and a scatter plot of the observed vs predicted probabilities with the line fit to that data and the line x=y added.