Causal Relationship between Stock market and Real Economy in India using Granger Causality test
Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a...
Transcript of Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a...
![Page 1: Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a useful paradigm for ANH research •It is important that we consider causality and](https://reader035.fdocuments.in/reader035/viewer/2022070112/6057a0192b2ea00ef1279c84/html5/thumbnails/1.jpg)
Probabilistic Causal Models for Nutrition Outcomes of
Agricultural Actions
Eike Luedeling
Cory Whitney
![Page 2: Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a useful paradigm for ANH research •It is important that we consider causality and](https://reader035.fdocuments.in/reader035/viewer/2022070112/6057a0192b2ea00ef1279c84/html5/thumbnails/2.jpg)
http://hdsa.org/wp-
content/themes/hdsa/images/img_livingHD_nutrition1.jpg
http://www.tatachemicals.com/upload
/images/masthead/Agriculture.jpg
How can we study the Agriculture-Nutrition linkage…?
![Page 3: Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a useful paradigm for ANH research •It is important that we consider causality and](https://reader035.fdocuments.in/reader035/viewer/2022070112/6057a0192b2ea00ef1279c84/html5/thumbnails/3.jpg)
http://hdsa.org/wp-
content/themes/hdsa/images/img_livingHD_nutrition1.jpg
http://www.tatachemicals.com/upload
/images/masthead/Agriculture.jpg
…if the link isn’t direct?
How can we study the Agriculture-Nutrition linkage…?
![Page 4: Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a useful paradigm for ANH research •It is important that we consider causality and](https://reader035.fdocuments.in/reader035/viewer/2022070112/6057a0192b2ea00ef1279c84/html5/thumbnails/4.jpg)
http://www.tatachemicals.com/upload
/images/masthead/Agriculture.jpg
http://hdsa.org/wp-
content/themes/hdsa/images/img_livingHD_nutrition1.jpg
…and involves many “confounding” factors?
…if the link isn’t direct?
How can we study the Agriculture-Nutrition linkage…?
![Page 5: Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a useful paradigm for ANH research •It is important that we consider causality and](https://reader035.fdocuments.in/reader035/viewer/2022070112/6057a0192b2ea00ef1279c84/html5/thumbnails/5.jpg)
The statistical challenge may not look like this:
![Page 6: Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a useful paradigm for ANH research •It is important that we consider causality and](https://reader035.fdocuments.in/reader035/viewer/2022070112/6057a0192b2ea00ef1279c84/html5/thumbnails/6.jpg)
…but like this:
A serious
challenge to our
common
statistical
procedures!
We’ll need deeper systems understanding to study this!
![Page 7: Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a useful paradigm for ANH research •It is important that we consider causality and](https://reader035.fdocuments.in/reader035/viewer/2022070112/6057a0192b2ea00ef1279c84/html5/thumbnails/7.jpg)
Confronting complexity
• A rather daunting challenge (?)
• Complete, ‘objective’ understanding may not be attainable (or prohibitively costly)
• Fortunately, for deciding whether an intervention works, we may not need to know everything…
• …and we can build on existing knowledge held by locals and experts
• Is existing knowledge sufficient for deciding whether a specific intervention works?
• If not, what do we need to measure?
Decision Analysis answers these questions
![Page 8: Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a useful paradigm for ANH research •It is important that we consider causality and](https://reader035.fdocuments.in/reader035/viewer/2022070112/6057a0192b2ea00ef1279c84/html5/thumbnails/8.jpg)
Causal decision model development with stakeholders
KenyaFruit trees on farms
UgandaVision 2040
Homegardens vs.
commercial farms
![Page 9: Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a useful paradigm for ANH research •It is important that we consider causality and](https://reader035.fdocuments.in/reader035/viewer/2022070112/6057a0192b2ea00ef1279c84/html5/thumbnails/9.jpg)
Decision Analysis
![Page 10: Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a useful paradigm for ANH research •It is important that we consider causality and](https://reader035.fdocuments.in/reader035/viewer/2022070112/6057a0192b2ea00ef1279c84/html5/thumbnails/10.jpg)
The Uganda Vision 2040 model
![Page 11: Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a useful paradigm for ANH research •It is important that we consider causality and](https://reader035.fdocuments.in/reader035/viewer/2022070112/6057a0192b2ea00ef1279c84/html5/thumbnails/11.jpg)
Conclusions
• Decision Analysis is a useful paradigm for ANH research
• It is important that we consider causality and adequately represent complexity
• Precise knowledge on complex agricultural systems can’t realistically be achieved at scale – fortunately it’s often not needed
• Decision Analysis provides strategies for using our state of knowledge (incl. uncertainty) to compare decision options and evaluate interventions
• Still work to do on adapting Decision Analysis approaches to work on agricultural development contexts
![Page 12: Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a useful paradigm for ANH research •It is important that we consider causality and](https://reader035.fdocuments.in/reader035/viewer/2022070112/6057a0192b2ea00ef1279c84/html5/thumbnails/12.jpg)
Our IMMANA experience
• Specific program to identify innovative approaches, which are urgently needed!
• Safe space to try new methods
• Fostered a necessary interdisciplinary conversation
• Incubator of innovation and ideas for future work
• Very well managed, incl. flexible where flexibility was needed
Thanks IMMANA!
![Page 13: Probabilistic Causal Models for Nutrition Outcomes of ... · Conclusions •Decision Analysis is a useful paradigm for ANH research •It is important that we consider causality and](https://reader035.fdocuments.in/reader035/viewer/2022070112/6057a0192b2ea00ef1279c84/html5/thumbnails/13.jpg)
Bayesian Network modeling
• Causal modeling approach
• Each node is related to its parents through conditional probability tables (if parent = A, then child = B)
• Each table only describes one relationship, which is usually easy to consider
• Probabilistic linkage
• Allows considering uncertainty and risk in a quantitative model
• Value of Information analysis, a capability added by this project, allows highlighting critical uncertainties