From 02aafb43d89be7e2e4079dc415e92855bc1edf42 Mon Sep 17 00:00:00 2001 From: Jakob Schumacher Date: Mon, 17 Jun 2024 21:06:37 +0200 Subject: [PATCH] Update 01-casual-to-causal.qmd In epidemiology rates are used only when the denominator is a measure of time. Which is usually not the case for prevalence. See https://www.thelancet.com/journals/laninf/article/PIIS1473-3099(20)30439-4/fulltext --- chapters/01-casual-to-causal.qmd | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/chapters/01-casual-to-causal.qmd b/chapters/01-casual-to-causal.qmd index 0061f9eb..3074e4d7 100644 --- a/chapters/01-casual-to-causal.qmd +++ b/chapters/01-casual-to-causal.qmd @@ -162,7 +162,7 @@ ggplot( Here are some other great examples of descriptive analyses. - Deforestation around the world. Our World in Data [@owidforestsanddeforestation] is a data journalism organization that produces thoughtful, usually descriptive reports on various topics. In this report, they present data visualizations of both absolute change in forest coverage (forest transitions) and relative change (deforestation or reforestation), using basic statistics and forestry theory to present helpful information about the state of forests over time. -- The prevalence of chlamydial and gonococcal infections [@Miller2004]. Measuring the prevalence of disease (how many people currently have a disease, usually expressed as a rate per number of people) is helpful for basic public health (resources, prevention, education) and scientific understanding. In this study, the authors conducted a complex survey meant to be representative of all high schools in the United States (the target population); they used survey weights to address a variety of factors related to their question, then calculated prevalence rates and other statistics. As we'll see, weights are helpful in causal inference for the same reason: targeting a particular population. That said, not all weighting techniques are causal in nature, and they were not here. +- The prevalence of chlamydial and gonococcal infections [@Miller2004]. Measuring the prevalence of disease (how many people currently have a disease, usually expressed as a ratio per number of people) is helpful for basic public health (resources, prevention, education) and scientific understanding. In this study, the authors conducted a complex survey meant to be representative of all high schools in the United States (the target population); they used survey weights to address a variety of factors related to their question, then calculated prevalence ratios and other statistics. As we'll see, weights are helpful in causal inference for the same reason: targeting a particular population. That said, not all weighting techniques are causal in nature, and they were not here. - Estimating race and ethnicity-specific hysterectomy inequalities [@Gartner2020]. Descriptive techniques also help us understand disparities in areas like economics and epidemiology. In this study, the authors asked: Does the risk of hysterectomy differ by racial or ethnic background? Although the analysis is stratified by a key variable, it's still descriptive. Another interesting aspect of this paper is the authors' work ensuring the research answered questions about the right target population. Their analysis combined several data sources to better estimate the true population prevalence (instead of the prevalence among those in hospitals, as commonly presented). They also adjusted for the prevalence of hysterectomy, e.g., they calculated the incidence (new case) rate only among those who could actually have a hysterectomy (e.g., they hadn't had one yet). #### Validity