Alessandro Rovetta
@AlessandroRov19
Followers
59
Following
220
Statuses
94
Co-founder of the International Committee Against the Misuse of Statistical Significance (ICAMSS). Soundtrack composer and co-founder of Redeev SRL.
Joined May 2020
Yes! We can use traditional "confidence" intervals as compatibility intervals! According to the chosen test, we can assess the compatibility (consistency) of the data with several effect hypotheses rather than the sole null hypothesis of zero effect. We thank @LancetRH_SEAsia!
P-values are often misinterpreted in medical research, leading to potential misjudgments about treatment effectiveness. A new Comment advocates for using compatibility intervals instead of relying solely on p-values. Read the Comment
0
5
32
What an amazing group, which includes Dr. Greenland and Dr. Ioannidis. Congrats @Mohamma70696197
I am very pleased to announce the launch of the “Journal of the Academy of Public Health”; my honor to be among such an impressive and esteemed group of Editorial Board members:
0
0
1
RT @Mohamma70696197: My 5th paper with Alessandro Rovetta on P-values and confidence intervals in medical research, just published in the L…
0
3
0
RT @VickersBiostats: There is no statistical solution to the problem of causal inference from observational data, it requires good scientif…
0
33
0
"Errors are randomly distributed" is a common assumption in statistics. However, in medicine, motivated reasoning and other biases often dramatically violate this expectation. Thank you @Lester_Domes, and @vk_wilde for sharing:
0
6
8
@VPrasadMDMPH I appreciate your work, but here, I think you are just confusing the 'absence of conclusive evidence' with the 'conclusive evidence of absence.' P>0.05 doesn't imply clinical non-significance; at best, it signals high uncertainty in the estimate according to your model.
0
0
0
@b1bb1d1b0bb1d1 @rnishimura I think it depends on what you mean: mathematical statistics is merely a set of rules for numerical calculation, whereas statistical science integrates the epistemic context into the analysis.
0
0
3
@AdanBecerraPhD @Mohamma70696197 [PART 2]. Therefore, I believe a fundamental step is to try to modify school curricula, starting from high school onwards. At the same time, it is important to continue raising awareness as much as possible to attract those who are genuinely interested in the issue.
0
0
1
@AdanBecerraPhD @Mohamma70696197 Dear Adan, Based on what I have studied so far, the misuse and abuse of statistical significance range from simple ignorance to conflicts of interest (e.g., it is easy to set and use α to reject inconvenient hypotheses and fail to reject convenient ones). [PART 1].
0
0
1
We published a method to analyze several hypotheses of clinical interest rather than the sole null hypothesis of an exactly zero effect (Journal of Preventive Medicine and Public Health, doi: 10.3961/jpmph.24.250). Accepted version: @Mohamma70696197
0
1
2
@vamrhein Dear Valentin, You might be interested in reading the comments from me and Dr. Mansournia. All the Best.
0
0
0
Don't dichotomize compatibility. It is wrong to say that hypotheses outside the 95%CI are incompatible with the data, while those within it are compatible. Support requires prior knowledge+bias active control. @Mohamma70696197
0
1
4
I don't know if they agree with this content, nor do I want to suggest that they do. Nevertheless, if there is anything right in what I have written here, I owe it to people like @Lester_Domes and @vamrhein
0
0
0
@vamrhein I love the idea, but some aspects need improvements IMHO. For instance, it is statistically wrong and infodemiologically dangerous to divide hypotheses into compatible and incompatible (this can lead to misuses and miscitations to support novel forms of dichotomy and P-hacking).
0
0
0
[STATISTICAL TESTING] I discuss why I think that chance is always the sole factor at play, even when dealing with non-null hypotheses. To summarize, this is because numbers and real phenomena are distinct concepts. This has practical implications.
0
0
2