Causal inference causal inference

Yüklə 1,45 Mb.

tarix	09.08.2018
ölçüsü	1,45 Mb.
	#62239

CAUSAL INFERENCE

A PROCESS OF PROVING THAT A PARTICULAR HYPOTHESIS IS A REAL ONE AND CAUSAL i.e., ESTABLISHING THE CAUSE AND EFFECT RELATIONSHIP BETWEEN A SUSPECTED FACTOR AND A DISEASE.
DIRECT CAUSAL ASSOCIATION MEANS THAT A FACTOR IS REALLY RELATED IN CAUSING THAT PARTICULAR DISEASE.
WHAT EVER AN ASOSIATION IS SEEN, ONE HAS TO PROVE ULTIMATELY WHETHER IT IS DIRECT CAUSAL ASSOSIATION OR NOT.

BY WEEDING OUT OR EXCLUDING VARIOUS FALSE LINKS AND REACHING THE REAL LINK

BY WEEDING OUT OR EXCLUDING VARIOUS FALSE LINKS AND REACHING THE REAL LINK
THIS PROCESS HAS TO BE DONE WITHOUT ANY ERRORS, EITHER SYSTEMATIC (BIASES) OR DUE TO CHANCE.

CAUSAL INFERENCE

A
SPECIAL CASE
OF

GENERAL PROCESS
OF
SCIENTIFIC REASONING

CONTRIBUTORS FOR SCIENTIFIC REASONING

17TH CENTURY – BACON-- INDUCTIVIST
18TH CENTURY – DAVID HUME -- DEDUCTIVIST
THOMAS BAYES-- MAKES THE SCIENTIST
RESPONSIBLE
19TH CENTURY - JAMES STUART MILL-- INDUCTIVIST
20TH CENTURY –
KARL POPPER -- REFUTATIONIST
KUHN, LAKTOS-- REJECTED KARL
SUSSER
21ST CENTURY – BRITISH ALL ARE USEFUL
PHILOSOPHERS & SCIENTIST JUDGEMENT
BETTER

INDUCTIVIST PHILOSOPHY (BACON & J.S.MILL)

INDUCTIVE REASONING BASED ON INTUSION THAT EACH EVENT IS FOLLOWED BY AN EFFECT (PRAGMATIC PHILOSOPHY)
OBSERVATIONS DRAWN FROM HYPOTHESES ARE CALLED INDUCTIONS
CONCLUSIONS ARE DRAWN FROM INDUCTIONS

J. S. MILL’S FIVE CANNONS (1856) METHODS OF INDUCTION

AGREEMENT: IF A FACTOR IS COMMON TO A NUMBER OF DIFFERENT CIRCUMSTANCES, THAT ARE ASSOCIATED WITH THE PRESENCE OF A DISEASE, THAT FACTOR MAY BE THE CAUSE OF DISEASE – THAT MEANS THERE IS AN AGREEMENT BETWEEN THE FACTOR AND THE DISEASE UNDER DIFFERENT CIRCUMSTANCES.
DIFFERENCE: IF THE FREQUENCY OF A DISEASE IS MARKEDLY DIFFERENT UNDER TWO DIFFERENT CIRCUMSTANCES AND SOME FACTORS CAN BE IDENTIFIED IN ONE CIRCUMSTANCE NOT IN OTHER, THEN THE FACTOR OR ITS ABSENCE, MAY BE THE CAUSE OF DISEASE.
CONCOMITANT VARIATION: FACTOR WHOSE FREQUENCY OR STRENGTH VARIES WITH THAT OF THE DISEASE, IT MAY BE THE CAUSE OF THE DISEASE.

HEALTH RESEARCH IS MAINLY EMPIRICAL RESEARCH AND LESS THEORITICAL.
HENCE IT DEPENDS ALMOST ENTIRELY ON INDUCTIVE REASONINING.

INDUCTIVE PHILOSOPHY

DEDUCTIVE LOGIC OF DAVID HUMES

DEDUCTIVISM

IN DEDUCTION , THE CONCLUSION NECESSARILY FOLLOWS FROM THE PREMISES ( All “A” is “B”, All “B” is “C”, therefore all “A” is “C”.).
DEDUCTION MOVES FROM GENERAL TO THE SPECIFIC AND DOES NOT ALLOW FOR THE ELEMENT OF CHANCE OR UNCERTAINITY.
DEDUCTIVE INFERENCES, THEREFORE, ARE SUITED TO THEORITICAL RESEARCH

KARL POPPER’S REFUTATIONISM OR FALSIFICATION THEORY

IF “H” IMPLIES “B” AND “B” IS FALSE, THEN “H” MUST BE FALSE
TRY TO DISPROVE , IF YOU CAN’T ACCECPT IT
EVEN 100 WHITE SWANS CANNOT PROVE THE HYPOTHESIS THAT “ALL SWANS ARE WHITE”. BUT JUST ONE NON –WHITE SWAN CAN DISPROVE IT.
THE CONDITION “IF” MAKES THE THEORY UNCERTAIN AND TENTATIVE.
SOUNDS PESSIMISTIC

BAYESIANISM

BAYES MAKES THE SCIENTIST RESPONSIBLE TO ASCERTAIN THE DEGREE OF CERTAINITY OR PERSONAL PROBABILITY TO THE ARGUMENT.
SCIENTIST HIMSELF HAS TO DECIDE ABOUT THE CERTAINITY BY ATTACHING HIS OWN (FROM HIS PERSONAL EXPERIENCE) PRIOR PROBABILITIES(INITIAL CERTAINITIES) AND POSTERIOR PROBABILITIES (CONCLUDING CERTAINITIES) EVENTHOUGH THEY ARE SUBJECTIVE.

TENTATIVENESS OF OUR KNOWLEDGE

All the fruits of any scientific work, epidemiological or of other disciplines, are at best only the tentative formulations of a description of nature. This tentativeness of our knowledge does not prevent us for practical applications but should keep us skeptical ad critical, not only of everyone else's work but our own as well

(oxford)

PRESENT VIEWS

Each philosophy has its own advantages and limitations.
Induction, deduction, falsification, scientist’s opinion – all are worth trying in appropriate circumstances with finer judgment.
Mills’ cannons are still often used in the forming of the hypotheses.

CAUSAL COMPONENTS

The causes other than necessary cause, complementing or helping the necessary cause are considered as components of the cause.
These are set of causes necessary to make a factor sufficient to cause the disease . They just complement each other and act as independent partners. They will not loose their individual identity, there is no change in their biological mechanisms.

Causal Complement:

Causal Complement:
Is entire set of factors or conditions and each component complement each other for causal co-action or joint action.
Sufficient cause:
It is a set of minimal (all necessary conditions and events) conditions and events that inevitably produce the disease. Sufficient cause competes causal mechanism and initiates the disease

CAUSAL INFERENCE EXCERCISE

Whenever an association is seen, one has to prove ultimately whether it is direct and real causal association or not. This process of proving is causal inference
All the non-causal statistical associations as well as spurious ones have to be eliminated
This has to be done without errors of selection, information, confounding and chance)

CHECK FOR BIASES

Is the observed association due to improper selection? Is it due to defective sampling? Is it due to the absence of comparison group?
Is it due imperfect measurement of the cause or its amount of exposure?re there any confounders like age, sex etc limiting the validity
Is it just due to chance?
After excluding all biases including that due chance, Hill’s checklist has to be applied to judge the causality with caution and proper judgment

HILL’S CHECKLIST FOR JUDGING THE CAUSALITY

TEMPORALITY
STRENGTH
SPECIFICITY
CONSISTANCY
COHERENCE
BIOLOGICAL PLAUSIBILITY
ANALOGY

SCOPE OF HILL’S CHECKLIST

Not hard and fast rules
Just additional qualities or aids to judge the causality
If any association is possessing these qualities, it is more in favor of causality but reverse is not true. An association even without these qualities may also be causal.
It is not necessary that all these qualities to be present as it is occurs very rarely.
Each of these qualities has to be looked for first, If present its weight should be judged individually and combined.
Even a few of them, if present, will help in causal judgment.

MAIN CONSIDERATIONS

Temporal sequence of the association i.e. whether the cause is preceding the effect or not, has to be searched first. If it is present, it is more in favor of causal association.
Then the strength of the association (the relative risk/ odd’s ratio and dose response relationship) which decides the power of the association between the cause and effect has to be determined. If relative risk is high, the association is more likely to be a causal one.
If temporality is present , a case control study (in urgent situations) or a cohort study (if time permits) may be initiated to find out relative risk/ odd’s ratio and to test for the causal association.

SUMMARY

Causal inference is an intelligent scientific interpretation exercise to know whether observed relationship is real or not and it is to be done without errors.
References :
Roger Detels, James Mc Even-Oxford Text
Book of Public Health
Brian Mac Mahan -Epidemiology-
principles & methods

Yüklə 1,45 Mb.

Dostları ilə paylaş:

Causal inference causal inference

CAUSAL INFERENCE

CAUSAL INFERENCE

CAUSAL INFERENCE

A PROCESS OF PROVING THAT A PARTICULAR HYPOTHESIS IS A REAL ONE AND CAUSAL i.e., ESTABLISHING THE CAUSE AND EFFECT RELATIONSHIP BETWEEN A SUSPECTED FACTOR AND A DISEASE.

DIRECT CAUSAL ASSOCIATION MEANS THAT A FACTOR IS REALLY RELATED IN CAUSING THAT PARTICULAR DISEASE.

WHAT EVER AN ASOSIATION IS SEEN, ONE HAS TO PROVE ULTIMATELY WHETHER IT IS DIRECT CAUSAL ASSOSIATION OR NOT.

BY WEEDING OUT OR EXCLUDING VARIOUS FALSE LINKS AND REACHING THE REAL LINK

BY WEEDING OUT OR EXCLUDING VARIOUS FALSE LINKS AND REACHING THE REAL LINK

THIS PROCESS HAS TO BE DONE WITHOUT ANY ERRORS, EITHER SYSTEMATIC (BIASES) OR DUE TO CHANCE.

CAUSAL INFERENCE

A

SPECIAL CASE

OF

CONTRIBUTORS FOR SCIENTIFIC REASONING

17TH CENTURY – BACON-- INDUCTIVIST

18TH CENTURY – DAVID HUME -- DEDUCTIVIST

THOMAS BAYES-- MAKES THE SCIENTIST

RESPONSIBLE

19TH CENTURY - JAMES STUART MILL-- INDUCTIVIST

20TH CENTURY –

KARL POPPER -- REFUTATIONIST

KUHN, LAKTOS-- REJECTED KARL

SUSSER

21ST CENTURY – BRITISH ALL ARE USEFUL

PHILOSOPHERS & SCIENTIST JUDGEMENT

BETTER

INDUCTIVIST PHILOSOPHY (BACON & J.S.MILL)

INDUCTIVE REASONING BASED ON INTUSION THAT EACH EVENT IS FOLLOWED BY AN EFFECT (PRAGMATIC PHILOSOPHY)

OBSERVATIONS DRAWN FROM HYPOTHESES ARE CALLED INDUCTIONS

CONCLUSIONS ARE DRAWN FROM INDUCTIONS

J. S. MILL’S FIVE CANNONS (1856) METHODS OF INDUCTION

AGREEMENT: IF A FACTOR IS COMMON TO A NUMBER OF DIFFERENT CIRCUMSTANCES, THAT ARE ASSOCIATED WITH THE PRESENCE OF A DISEASE, THAT FACTOR MAY BE THE CAUSE OF DISEASE – THAT MEANS THERE IS AN AGREEMENT BETWEEN THE FACTOR AND THE DISEASE UNDER DIFFERENT CIRCUMSTANCES.

DIFFERENCE: IF THE FREQUENCY OF A DISEASE IS MARKEDLY DIFFERENT UNDER TWO DIFFERENT CIRCUMSTANCES AND SOME FACTORS CAN BE IDENTIFIED IN ONE CIRCUMSTANCE NOT IN OTHER, THEN THE FACTOR OR ITS ABSENCE, MAY BE THE CAUSE OF DISEASE.

CONCOMITANT VARIATION: FACTOR WHOSE FREQUENCY OR STRENGTH VARIES WITH THAT OF THE DISEASE, IT MAY BE THE CAUSE OF THE DISEASE.

HEALTH RESEARCH IS MAINLY EMPIRICAL RESEARCH AND LESS THEORITICAL.

HENCE IT DEPENDS ALMOST ENTIRELY ON INDUCTIVE REASONINING.

INDUCTIVE PHILOSOPHY

DEDUCTIVE LOGIC OF DAVID HUMES

DEDUCTIVISM

IN DEDUCTION , THE CONCLUSION NECESSARILY FOLLOWS FROM THE PREMISES ( All “A” is “B”, All “B” is “C”, therefore all “A” is “C”.).

DEDUCTION MOVES FROM GENERAL TO THE SPECIFIC AND DOES NOT ALLOW FOR THE ELEMENT OF CHANCE OR UNCERTAINITY.

DEDUCTIVE INFERENCES, THEREFORE, ARE SUITED TO THEORITICAL RESEARCH

KARL POPPER’S REFUTATIONISM OR FALSIFICATION THEORY

IF “H” IMPLIES “B” AND “B” IS FALSE, THEN “H” MUST BE FALSE

TRY TO DISPROVE , IF YOU CAN’T ACCECPT IT

EVEN 100 WHITE SWANS CANNOT PROVE THE HYPOTHESIS THAT “ALL SWANS ARE WHITE”. BUT JUST ONE NON –WHITE SWAN CAN DISPROVE IT.

THE CONDITION “IF” MAKES THE THEORY UNCERTAIN AND TENTATIVE.

SOUNDS PESSIMISTIC

BAYESIANISM

BAYES MAKES THE SCIENTIST RESPONSIBLE TO ASCERTAIN THE DEGREE OF CERTAINITY OR PERSONAL PROBABILITY TO THE ARGUMENT.

SCIENTIST HIMSELF HAS TO DECIDE ABOUT THE CERTAINITY BY ATTACHING HIS OWN (FROM HIS PERSONAL EXPERIENCE) PRIOR PROBABILITIES(INITIAL CERTAINITIES) AND POSTERIOR PROBABILITIES (CONCLUDING CERTAINITIES) EVENTHOUGH THEY ARE SUBJECTIVE.

TENTATIVENESS OF OUR KNOWLEDGE

PRESENT VIEWS

Each philosophy has its own advantages and limitations.

Induction, deduction, falsification, scientist’s opinion – all are worth trying in appropriate circumstances with finer judgment.

Mills’ cannons are still often used in the forming of the hypotheses.

CAUSAL COMPONENTS

The causes other than necessary cause, complementing or helping the necessary cause are considered as components of the cause.

These are set of causes necessary to make a factor sufficient to cause the disease . They just complement each other and act as independent partners. They will not loose their individual identity, there is no change in their biological mechanisms.

Causal Complement:

Causal Complement:

Is entire set of factors or conditions and each component complement each other for causal co-action or joint action.

Sufficient cause:

It is a set of minimal (all necessary conditions and events) conditions and events that inevitably produce the disease. Sufficient cause competes causal mechanism and initiates the disease

CAUSAL INFERENCE EXCERCISE

Whenever an association is seen, one has to prove ultimately whether it is direct and real causal association or not. This process of proving is causal inference

All the non-causal statistical associations as well as spurious ones have to be eliminated

This has to be done without errors of selection, information, confounding and chance)

CHECK FOR BIASES

Is the observed association due to improper selection? Is it due to defective sampling? Is it due to the absence of comparison group?

Is it due imperfect measurement of the cause or its amount of exposure?re there any confounders like age, sex etc limiting the validity

Is it just due to chance?

After excluding all biases including that due chance, Hill’s checklist has to be applied to judge the causality with caution and proper judgment

HILL’S CHECKLIST FOR JUDGING THE CAUSALITY

TEMPORALITY

STRENGTH

SPECIFICITY

CONSISTANCY

COHERENCE