Dr Mircea Zloteanu 🌺🌞🍃#statstab #535 The Benefits Of Reporting Critical Effect Size Values
Thoughts: You ran a study w/o a power analysis? Report the lowest value your test could have detected at 80% power and 5% alpha.
skua@ki_sekiya
2/2
Not convinced by the stats in the article. I thought that a random sample of +70 ispections would usually provide accurate reliable conclusions.
".. the Clean Energy Regulator carried out 1,278 compliance inspections.
...
The sample size in the regulator’s report is small – 0.5% of the total number of systems installed.
With such a small sample size, i
t is hard to extrapolate the level of installation non-compliance across all systems in Australia."
#Statistics #SampleSize
Jan R. BoehnkeAnother #PeerReview done.
Manuscript c3,000 words
Review c2,300 words
3.25hrs
I do love Null results.*
Nevertheless, a good theoretical background is important (and ideally written down before the results are known).
It should be clear what an effect could look like.
#EffectSize #SampleSize
Superiority is different to non-inferiority.
#RCT
#PreRegistration #RegisteredReport
* ad libbing on Julia Rohrer's post here:
https://www.the100.ci/2017/06/01/why-we-should-love-null-results/
Why we should love null results
or Dear Sanjay TL;DR: Publication bias is a bitch, but poor hypothesising may be worse. Estimated reading time: 10 minutes A few weeks ago I was listening to episode 5 of the Black Goat, flowery thoughts on my mind, when suddenly I heard Sanjay Srivastava say the following words (from minu
Statistics GlobeToday, I released a new module in the Statistics Globe Hub that explains how to perform sample size calculation using power analysis.
More info about the Statistics Globe Hub: https://statisticsglobe.com/hub
#rstats #datascience #statistics #poweranalysis #samplesize #studyplanning #statisticsglobehub
#statstab #492 A tiny review on e-values and e-processes
Thoughts: How to be a bayesian while wearing a frequentist hat.
#evalues #eprocess #samplesize #evidence #power #sequential #error #type1
Emotional_Series7814
#statstab #489 On the performance of the Neyman Allocation with small pilots
Thoughts: If you know your treatment condition will have larger variance you can optimise your sample size.
#nhst #samplesize #neynan #heterogeneity #welch #variance #pilot #se
https://www.sciencedirect.com/science/article/pii/S0304407624001398
#statstab #477 Don’t calculate post-hoc power using observed estimate of effect size
Thoughts: Good discussion and many useful references. Even big journals print stupid stuff.
#posthoc #power #sensitivity #samplesize #consort #medicine #bias
#statstab #470 Low power bias {shiny} app
Thoughts: easily show what conducting an underpowered study would do to your effect size (type M error).
#teaching #bias #power #typeM #typeS #QRPs #underpowered #samplesize
Another #PeerReview done.
Manuscript c4,000 words
Review c2,700 words
5hrs
Paper in a key area of my methodological work, so it was really interesting. But I really needed to get stuck in.
Two collaboration projects on the design and reporting of #RCTs that might be useful for others:
https://pubmed.ncbi.nlm.nih.gov/37982521/
presents 19 factors to aid trial design, and the DELTA2 Guidance specifying a target difference and reporting the #SampleSize calculation for RCTs
https://trialsjournal.biomedcentral.com/articles/10.1186/s13063-018-2884-0
Appropriate design and reporting of superiority, equivalence and non-inferiority clinical trials incorporating a benefit-risk assessment: the BRAINS study including expert workshop - PubMed
Funded by the Medical Research Council UK and the National Institute for Health and Care Research as part of the Medical Research Council-National Institute for Health and Care Research Methodology Research programme.