I try to put straight lines through things but usually fail. Try to be Bayesian when I can. Views my own. RT/like != endorsement. Graduate of YouTube University
Interesting paper! Thanks for posting. IIUC, your main concern with the GAM approach is that it targets the wrong estimand. If so, I feel that your criticism of the approach is kind of unfair, given that it's easy to target CME w/ GAM. See this notebook: arelbundock.com/hmx_simonsoh…
Large boulder the size of a small boulder is completely blocking east-bound lane Highway 145 mm78 at Silverpick Rd. Please use caution and watch for emergency vehicles in the area.
Check out the latest working paper by Prof. @dlmillimet, entitled
"Fixed Effects and Causal Inference"
available through @iza_bonn. #PonyUpdocs.iza.org/dp16202.pdf
ALT Text from https://www.utstat.toronto.edu/brunner/oldclass/378f16/readings/CohenPower.pdf, pg 79 to 80.
The key point is that we can get r = .1 from a true correlation of .25 when there is non-differential measurement error that attenuates the true relationship.
Okay, psychologists. When did you collectively decide that an effect size of r = .1 is the cutoff for meaningful effect sizes?
Source: sciencedirect.com/science/ar…
ALT Most relationships between digital media use and well-being identified in this study were weak (below 0.1). If we set above 0.1 as a meaningful level for the effects of digital media, anything below 0.1 can be considered too minor to be meaningful. From the perspective of confidence intervals the confidence intervals of texting and SNS on overall well-being and the effect sizes of most SNS activities on overall well-being were completely contained in the null region (−0.10 to 0.10) and could be considered null (or at best, trivial) correlations. When excluding these null results from our analyses, SNS use, including both general use and specific activities, had more meaningful associations with psychopathology than with positive well-being. Among all types of digital media use, only the results of phone calls (for positive well-being), SNS use (for psychopathology), and online gaming (for psychopathology) were meaningful.
Okay, psychologists. When did you collectively decide that an effect size of r = .1 is the cutoff for meaningful effect sizes?
Source: sciencedirect.com/science/ar…
ALT Most relationships between digital media use and well-being identified in this study were weak (below 0.1). If we set above 0.1 as a meaningful level for the effects of digital media, anything below 0.1 can be considered too minor to be meaningful. From the perspective of confidence intervals the confidence intervals of texting and SNS on overall well-being and the effect sizes of most SNS activities on overall well-being were completely contained in the null region (−0.10 to 0.10) and could be considered null (or at best, trivial) correlations. When excluding these null results from our analyses, SNS use, including both general use and specific activities, had more meaningful associations with psychopathology than with positive well-being. Among all types of digital media use, only the results of phone calls (for positive well-being), SNS use (for psychopathology), and online gaming (for psychopathology) were meaningful.
Chasing Shadows: How Implausible Assumptions Skew Our Understanding of Causal Estimands. Stijn Vansteelandt & Kelly Van Lancker. Statistics in Biopharmaceutical Research. tandfonline.com/doi/full/10.…
And another post on causal inference and social media and its effect on mental health. DAGs make a prominent appearance.
As always, feedback welcome (especially if you think I am wrong).
sjwild.github.io/blog/2024/1…
DAGs can be used to understand the mechanics of linear interaction analysis. Easy to see the zero correlation between the first-order and interaction terms after centering (though this is not the reason for centering). See more here: bpspsychub.onlinelibrary.wil…