@Iplaywithgerms This paper gives documentation on software (with causal reasoning, assumptions reviewed in appendix) for a parametric approach to estimating either "total effects" or "controlled direct effects" with competing events and time-varying

@JuliaLMarcus @Iplaywithgerms Total effects capture paths by which treatment affects competing event (e.g. protective total effect of lifesaving treatment on dementia may be wholly/partially due to effect on survival). Controlled direct effects do not capture these paths
@JuliaLMarcus @Iplaywithgerms More detailed reasoning on the difference and tradeoffs between total and controlled direct effects and causal reasoning in the point treatment context provided here along with description of some estimators and code.
https://t.co/uRQcU1NqJd
@JuliaLMarcus @Iplaywithgerms If you are familiar with more robust approaches like IPW or even better TMLE for time-varying treatment, these are trivially adapted to go after the controlled direct effect by simply treating competing events like loss to follow-up (censoring). e.g. see
https://t.co/kew3HltAev
@JuliaLMarcus @Iplaywithgerms Examples of IPW estimation of the total effect of a time-varying treatment described in Appendix D of this paper:
https://t.co/RNhcgTBMkb
And here
https://t.co/rMWmwFBWwV
Others in reference lists of above papers.
@JuliaLMarcus @Iplaywithgerms As discussed here
https://t.co/uRQcU1NqJd
even when treatment is time-fixed, hazard ratios (whether cause-specific or subdistribution) do not quantify causal effects (even when counterfactual contrasts) except in special circumstances that will not hold in most studies
@JuliaLMarcus @Iplaywithgerms These circumstances can fail even when the study is a perfectly executed trial (no loss to follow-up, perfect adherence to the treatment assignment) and even when the treatment does not affect the competing event.
@JuliaLMarcus @Iplaywithgerms More recent work on the separable effects provides notions of mechanism in this setting that, unlike controlled direct effects (effects under eliminating competing events), have real-world, clinical relevance. This does not accommodate t-v treatments (yet, in the works).
@JuliaLMarcus @Iplaywithgerms Here are some papers on the separable effects for point treatments (Epidemiology commentary forthcoming focused on fertility trials that is less technical):
https://t.co/35OJDe8dD6
https://t.co/O08FNqPWWb
https://t.co/wKCrItvNhT

More from Software

Are you a Designer or a Developer?👨‍💻

Here are some Google Chrome extensions that can make you better in 2021. 🔥🍀

(Thread) 🧵👇

1.
https://t.co/zGir5E5U0J: https://t.co/PVx1wlX0Se is the easiest way to stay updated on the latest programming news. Get the hottest dev news from the best tech blogs on any topic you can think of.

2. CSS Peeper: CSS Peeper is a CSS viewer tailored for Designers. Get access to useful styles with our Chrome extension. Its mission is to let Designers focus on design, and spend as little time as possible digging in a

3. UX Check: UX Check makes heuristic evaluations quick and easy. The extension will open up Nielsen's Ten Heuristics in a side pane next to your website.

4. Checkbot: Checkbot finds critical SEO, speed & security problems before your website visitors do
Tests 100s of pages at once for broken links, duplicate titles, invalid HTML, insecure pages, and 50+ other
Kubernetes vs Serverless offerings

Why would you need Kubernetes when there are offerings like Vercel, Netlify, or AWS Lambda/Amplify that basically manage everything for you and offer even more?

Well, let's try to look at both approaches and draw our own conclusions!

🧵⏬

1️⃣ A quick look at Kubernetes

Kubernetes is a container orchestrator and thus needs containers to begin with. It's a paradigm shift to more traditional software development, where components are developed, and then deployed to bare metal machines or VMs.

There are additional steps now: Making sure your application is suited to be containerized (12-factor apps, I look at you:
https://t.co/nuH4dmpUmf), containerizing the application, following some pretty well-proven standards, and then pushing the image to a registry.

After all that, you need to write specs which instruct Kubernetes what the desired state of your application is, and finally let Kubernetes do its work. It's certainly not a NoOps platform, as you'll still need people knowing what they do and how to handle Kubernetes.



2️⃣ A quick look at (some!) serverless offerings

The offer is pretty simple: You write the code, the platform handles everything else for you. It's basically leaning far to the NoOps side. There is not much to manage anymore.

Take your Next.js / Nuxt.js app, point the ...
The Great Software Stagnation is real, but we have to understand it to fight it. The CAUSE of the TGSS is not "teh interwebs". The cause is the "direct manipulation" paradigm : the "worst idea in computer science" \1


Progress in CS comes from discovering ever more abstract and expressive languages to tell the computer to do something. But replacing "tell the computer to do something in language" with "do it yourself using these gestures" halts that progress. \2

Stagnation started in the 1970s after the first GUIs were invented. Every genre of software that gives users a "friendly" GUI interface, effectively freezes progress at that level of abstraction / expressivity. Because we can never abandon old direct manipulation metaphors \3

The 1990s were simply the point when most people in the world finally got access to a personal computer with a GUI. So that's where we see most of the ideas frozen. \4

It's no surprise that the improvements @jonathoda cites, that are still taking place are improvements in textual representation : \5

You May Also Like