Utilizing and understanding numbers as a part of wellbeing news and examination

2660 days ago, 894 views
PowerPoint PPT Presentation
Utilizing and comprehension numbers as a part of wellbeing news and examination ...

Presentation Transcript

Slide 1

Utilizing and understanding numbers as a part of wellbeing news and research Heejung Bang, PhD Department of Public Health Weill Medical College of Cornell University

Slide 2

A basis for now's discussion Coffee is terrible yesterday, however great today and awful again tomorrow. "It's the cure of the week or the enemy of the week, the risk of the week." says B. Kramer. "I've seen such a large number of conflicting studies with espresso that I've come to overlook all of them." says D. Berry. What to accept? For some time, you may simply continue drinking espresso.

Slide 3

Hardly a day passes by without another feature about the assumed wellbeing dangers or advantages of something… Are these features advocated? Frequently, the answer is NO .

Slide 4

R. Peto phrases the way of the contention thusly: " Epidemiology is so excellent and gives such an imperative point of view on human life and passing, yet a mind blowing measure of junk is distributed ," by which he implies the aftereffects of observational studies that seem day by day in the news media and frequently turn into the premise of general wellbeing proposals about what we ought to or ought not do.

Slide 5

3 noteworthy explanations behind espresso like circumstances Confounding Multiple testing Faulty outline/test choice

Slide 6

Topics to be secured today Numbers in public statement Lies, Damn Lies & Statistics Association versus Causation Experiment (e.g., RCT) versus Observational study Replicate or Perish Hierarchy of proof and study plan Meta-investigation Multiple testing Same words, diverse implications? Information sharing Other Take-Home messages

Slide 7

1. Numbers in official statement No p-esteem, no chances or perils proportion in public statement! - Ask individuals in the city "what is p-esteem?" - Only we may chuckle in the event that I make a factual joke utilizing 0.05, 1.96 and 95%, and so on

Slide 8

What is P-esteem? In factual speculation testing , the p-esteem is the likelihood of getting an outcome in any event as extraordinary as a given information point, under the invalid theory . - If there is no speculation, there is no test and no p-esteem. Current factual preparing and practice, measurable testing/p-esteem are excessively underlined. Be that as it may, p-esteem (1 number, 0-1) can be helpful to basic leadership. - you can't say "it depends" every one of the times despite the fact that it can be valid.

Slide 9

Numerator & denominator Always attempt to check numerator and denominator (and when, to what extent) Try to peruse commentaries under * - 100% expansion can be 1 → 2 cases - 20% occasion rate can be 1 out of 5 tests

Slide 10

Large Number myths With vast N, one will more probable discover a distinction when a distinction really exists – idea of measurable power. In any case, numerous crucial issues (e.g., inclination, frustrating and wrong specimen choice) CANNOT be cured by expansive N. (all the more later) Combining numerous off base stories can make more major issues than reporting a solitary mistaken story. (all the more later in meta) N>200,000 expected to identify 20% decrease in mortality (Mann, Science 1990) Means (and t-test) can be extremely unsafe b/c with huge N, everything is huge - Perhaps, for DNA and race, Watson ought to see the whole appropriation or SD!

Slide 11

2. Lies, condemned lies & insights There are three sorts of untruths - B Disraeli & M Twain - Title justifies itself with real evidence "J Robins makes measurements come clean: Numbers in the administration of wellbeing" (Harvard Gazette meet) If numbers/insights are legitimately created and utilized, they can be the best bit of observational confirmation . - some experimental proof is quite often great to have - it is difficult to battle with numbers (and age)!

Slide 12

Some Advice No insights is superior to awful measurements. Simply present your information (e.g., N=3) when insights are a bit much. Enlightening insights versus inferential insights If you utilize wrong details, you can be on the news. See 'Factual imperfection trips up investigation of awful details'. Nature 2006

Slide 13

3. Affiliation versus Causation #1 mistake in wellbeing news, Association=Causation In 1748, D. Hume expressed 'we may characterize a cause to be a protest took after by another… where, if the principal question had not been, the second never had existed.' - this is a genuine cause! A more significant quote from Hume is " All contentions concerning presence are established on the connection of circumstances and end results .'

Slide 14

Misuses and misuse of "causes" You may stay away from the words 'cause', 'dependable', 'impact', "effect" or "impact" in your paper or public statement (esp., title), if results are gotten from observational studies. Rather you may utilize "affiliation" or 'relationship'. Regularly, "may/might" insufficient. Media abuses and open misconstrues this extremely - Every morning, we hear new reasons for some ailment are found.

Slide 15

half hazard decrease, 20% hazard diminishment, etc. In the event that you include, at this point all reasons for tumor (& numerous different infections) ought to have been recognized. All are affiliation, not causation. - there are an exceedingly extensive number of related and connected elements, contrasted with genuine causes. - a review of 246 recommended coronary hazard elements. Hopkins & Williams (1981) - I trust disease >1000 chance variables. 'Excessively numerous don't do' is no superior to 'do anything'.

Slide 16

Why Association ��  Causation? Confounders otherwise known as, third variable(s) Biggest risk to any observational studies. Meaning of 'frustrate': vt. Toss (things) into turmoil; stir up; confound. (Oxford Dictionary) However, confounders CANNOT be characterized as far as factual ideas alone (Pearl)

Slide 17

Confounder samplers Gray Hair versus heart assault Stork versus birth rate Rock & Roll versus HIV Eating late & weight pick up? Drinking (or match-conveying) & lung malignancy No father's name & newborn child mortality Long leg & skin disease Vitamins/HRT, as well? Any cure? - first thing to do is 'Utilize sound judgment'. Consider whatever other (concealed) variable or option clarification'.

Slide 18

Common sense & luck Common sense is the reason for the vast majority of the thoughts for planning logical examinations. - M Davidian in spite of the fact that we ought not overlook the significance of good fortune in science

Slide 19

By the way, why "causes" are so critical? On the off chance that causes can be expelled, powerlessness stops to matter (Rose 1985) and the result won't happen. Neither related nor connected elements have this power. Readily, a few endeavors have been made: 'Recognizing Association from Causation: A Backgrounder for Journalists' from American Council on Science and Health

Slide 20

Greenland's Dictum (Science 1995) There is nothing wicked about going out and getting proof, such as asking individuals what amount do you drink and checking bosom tumor records. There's nothing evil about checking whether that proof corresponds . There's nothing corrupt about checking for bewildering factors. The transgression comes in trusting a causal theory is genuine in light of the fact that your study thought of a positive result, or trusting the inverse on the grounds that your study was negative.

Slide 21

Association to causation? In 1965, Hill proposed an arrangement of the accompanying causal criteria: Strength Consistency Specificity Temporality (i.e., cause before impact) Biological inclination Plausibility Coherence Experiment Analogy However, Hill likewise said "None of my nine perspectives can bring undeniable confirmation for or against the circumstances and end results theory and none can be required as a sine qua non '.

Slide 22

Another enormous issue: predisposition and defective outline/tests Selection inclination: the contortion of a measurable examination, because of the technique for gathering tests. The most effortless approach to cheat (purposefully or unexpectedly) - Make group1: group2 = solid individuals: wiped out individuals. - Oftentimes, treatment is terrible in observational studies, why? - Do a study among your companions just - People are not the same as the starting?? (e.g., veggie lovers versus meat-significant other, HRT clients versus non-clients) Case-control concentrate on & coordinating: simple to say yet difficult to do accurately. - Vitamin C and tumor For any examination: FAIRNESS is generally critical! - Numerous different predispositions exist

Slide 23

Would you trust these p-values? (Cameron and Pauling, 1976) This well known study has neglected to reproduce 16 or so times! Pauling got two Nobel.

Slide 24

4. Analyze versus Observational study Although the contending from trials and perceptions by enlistment be no exhibition of general conclusions, yet it is the most ideal method for belligerence which the way of things concedes to. - I Newton's "experimental philosophy" of science: Science ought not, as Descartes contended, be founded on crucial standards found by reason, however in view of key sayings appeared to be valid by analyses.

Slide 25

Why clinical trials are vital? Randomized Controlled Trial (RCT) is the most widely recognized type of investigation on people. 'Normal causal impacts' can be evaluated from trial. - To know the genuine impact of treatment inside individual, one ought to be dealt with and untreated in the meantime. Experimentation trumps perception. (force of coin-flip! Confounders vanish.) Very hard to cheat in RCTs (because of randomization and convention). " Causality : God knows however people require a time machine. At the point when God is occupied and no time machine is accessible, a RCT would do ."

Slide 26

Problems/issues of RCTs Restrictive settings Human subjects under analyses Can be exploitative or infeasible Short terms 1-2 medicines, 1-2 dosages just Limited generalizability Other issues: blinding, drop-up, consistence

Slide 27

Problems/issues of observational studies Bias & jumbling Post-hoc contentions about natural credibility must be seen with some suspicion since the human creative energy appears to be fit for building up a rationa