Bayesian statistics is gaining popularity in medical science particularly after covid. Which approach to use? Which approach to be believed in treatment decisions? What approach is better for evidence synthesis? -A thread.@nirajvasavada Well, what approach to use and believe depends upon one important question. What you seek? Or what is your research question. Let us understand via an example given in a book by author Ben Lambert.

Bayesian statistics- Ben lambert

The Purpose of Statistical Inference. @nirajvasavada How much does a particular drug affect a patient’s condition? What can an average student earn after obtaining a college education? In life, we develop theories and use these to make predictions, but testing those theories is not easy.

We use a Receiver Operator Characteristic (ROC) curve in many research studies. What it is actually? And what does it signifies.@nirajvasavada A ROC curve is constructed by plotting the true positive rate (TPR) against the false positive rate (FPR). TPR is nothing be sensitivity and FPR is actually 1-specificity

Why do we need to define the primary outcome in any research? Why studying multiple outcomes is not desirable from the same study.? What is Bonferroni correction of p-value. @nirajvasavada Generally in any research, we calculate sample size according to primary outcome and try to keep false positive less than 5% (type 1 or alpha error 0.05) and false-negative less than 20% (type 2 or beta error)

Is significant results in large sample size always good or do insignificant result always mean lack of effect ? why we need to undestand effect size? A curious case of molnupiravir in vaccinated people vs corticosteriods in alchoholic hepatitis. @nirajvasavada @nirajvasavada We do not require statistics for knowing that an elephant is larger than an ant because that is true in 100% case.

Every medical practitioners reading research or speaking in conference or implementing in practice should know the following in simpler terms. (I am not going in to details) @nirajvasavada In large enough samples sometimes smaller difference in groups can give significant p value, p value is not measure of strength or how effective the treatment is it is just showing that roughly high chance that there is difference in the group. (Or your bill hypothesis is false)

2-deoxy s glucose is another talk of the town. It inhibit glycolysis and causes death of highly glycolytic cells like macrophages which also promote viral replication and hence viral infected cells die.@nirajvasavada The data (press note) available had low sample size with limited benefit on hospital stay. (Who score), high chances of false positive and negative results. So adjuvant therapy with little effect.

“It is a natural progression in any profession,for the student to become the teacher, the mentee to become the mentor.
It’s not that much different from parenting. We learn from our ancestors how to make the perfect soup.How to sooth an aching head or heart.We measure milestones, percentiles and first steps, to try and gain insight in to the person they will grow into.

@PypAyurved @omlakhani I am not going in debate regarding paid and non paid and I encourage such efforts and hope Ayurvedic medicine feels the pressure and come with more studies. Having said that this article has many flaws and with this article (first link) nothing is proved as under.@nirajvasavada @PypAyurved @omlakhani @nirajvasavada First sample size is very small and non calculated and 50 and 45 patient in each group gives this article no statical power (1-beta error) so very very high chance of false positive results.