updated readme.

README.Rmd

@@ -28,9 +28,9 @@ knitr::opts_chunk$set(collapse = TRUE, comment = "#>",
 
 Using predictions from pre-trained algorithms as outcomes in downstream statistical analyses can lead to biased estimates and misleading conclusions. The statistical challenges encountered when drawing inference on predicted data (IPD) include:
 
-1.  Understanding the relationship between predicted outcomes and their true, unobserved counterparts
-2.  Quantifying the robustness of the AI/ML models to resampling or uncertainty about the training data
-3.  Appropriately propagating both bias and uncertainty from predictions into downstream inferential tasks
+1.  Understanding the relationship between predicted outcomes and their true, unobserved counterparts.
+2.  Quantifying the robustness of the AI/ML models to resampling or uncertainty about the training data.
+3.  Appropriately propagating both bias and uncertainty from predictions into downstream inferential tasks.
 
 Several works have proposed methods for IPD, including post-prediction inference (PostPI) by [Wang et al., 2020](https://www.pnas.org/doi/suppl/10.1073/pnas.2001238117), prediction-powered inference (PPI) and PPI++ by [Angelopoulos et al., 2023a](https://www.science.org/doi/10.1126/science.adi6000) and [Angelopoulos et al., 2023b](https://arxiv.org/abs/2311.01453), and post-prediction adaptive inference (PSPA) by [Miao et al., 2023](https://arxiv.org/abs/2311.14220). Each method was developed to perform inference on a quantity such as the outcome mean or quantile, or a regression coefficient, when we have:
 
@@ -192,8 +192,8 @@ fig1
 
 We can see that:
 
--   The predicted outcomes are more correlated with the covariate than the true outcomes (plot A)
--   The predicted outcomes are not perfect substitutes for the true outcomes (plot B)
+-   The predicted outcomes are more correlated with the covariate than the true outcomes (plot A).
+-   The predicted outcomes are not perfect substitutes for the true outcomes (plot B).
 
 ### Model Fitting
 

README.md

@@ -30,11 +30,11 @@ conclusions. The statistical challenges encountered when drawing
 inference on predicted data (IPD) include:
 
 1.  Understanding the relationship between predicted outcomes and their
-    true, unobserved counterparts
+    true, unobserved counterparts.
 2.  Quantifying the robustness of the AI/ML models to resampling or
-    uncertainty about the training data
+    uncertainty about the training data.
 3.  Appropriately propagating both bias and uncertainty from predictions
-    into downstream inferential tasks
+    into downstream inferential tasks.
 
 Several works have proposed methods for IPD, including post-prediction
 inference (PostPI) by [Wang et al.,
@@ -160,9 +160,9 @@ relationships between these variables:
 We can see that:
 
 - The predicted outcomes are more correlated with the covariate than the
-  true outcomes (plot A)
+  true outcomes (plot A).
 - The predicted outcomes are not perfect substitutes for the true
-  outcomes (plot B)
+  outcomes (plot B).
 
 ### Model Fitting