Saturday, September 9, 2023

Selection bias in real world academic intervention trials: hybrid human AI conversational learning

 Summary : Real world academic intervention trials go through a rigorous selection process that could be possibly fraught with bias. Hybrid human AI conversational learning transcripts below toward resolution of bias. 



 [9/8, 8:06 PM] Dr Anil : Government Approved Fees Of Private medical College in Rajasthan.
 
MD RADIO-3.6 CRORE
MD DERMAT- 3.3 CRORE
MD MEDICINE -2.4 CRORE
MS OBGY- 2.4 CRORE
MD PAED- 1.95 Cr
MD RESP MEDICINE -1.8 CRORE
MS ORTHO- 1.65Cr

Note-This University is not even a Deemed University


[9/8, 8:11 PM] Thanga Prabhu: Scam


[9/8, 8:12 PM] Thanga Prabhu: Not worth at all to even think of spending this kind of money. I will keep away from being treated by anyone who has gone down this route. We need a de novo definition of Healer.


[9/8, 8:13 PM] Thanga Prabhu: There are better ways to specialize and be a healer


[9/8, 8:13 PM] Sundar clinical engineer : Rationalising the education fees should be taken up hand in hand with rationalising treatment fees.


[9/8, 8:16 PM] Thanga Prabhu: Politicians own most med colleges. It is foolish to expect them to cull a hen laying eggs in crores. We have to move away from such scams. When demand drops, like engg seats these also will lose value


[9/8, 8:26 PM] Dr Anil : So the blame automatically goes to Politicians and NO self reflection from community,NO filling cases against MCI,Nothing to be done as a self healing practice by the community...🤐

9/9, 8:19 AM] Rakesh Biswas: Only 80% of the community is sheep while the rest 20% are wolves (politicians) in sheep's clothing!

[9/9, 8:20 AM] Sundar Pulmo : Wolves are there everywhere 😅


[9/9, 8:21 AM] Metapsychist Number 1 Kims 2015: Every voter has to be a politician in democracy.
If they don't play politics,politics play them.

Can play healthy politics too


[9/9, 8:23 AM] Rakesh Biswas: The voter (sheep prey) has to now become a predator (in sheep's clothing)?


[9/9, 8:24 AM] Rakesh Biswas: Healthy as in :

You don't eat me and I don't eat you! Let's all eat grass?



[9/8, 8:26 PM] Sundar CE : True that. But then politicians also own hospitals and it would pinch them to see the fees being reduced, no?

[9/8, 8:31 PM] Sundar CE : Or, the MCI liberalises the permission to open medical colleges that can share hospital facilities between them and have less faculty than it is today. I think Dr Devi Shetty had advocated this some time back. If this is a good idea, we will soon have surplus seats and prices will fall down.



[9/8, 8:32 PM] Sundar Clinical Engineer: https://youtube.com/watch?v=xewwXKqmCyY&si=6NRlC0LzBQV02nmW


[9/8, 10:50 PM] Thanga Prabhu: Faculty shortage is real but reducing them or overloading them with extra responsibilities is retrograde step. Tech plays big part in force multiplying existing faculty. AI and haptics alongwith gaming tech is already being used across our med colleges.


[9/8, 11:11 PM] Metapsychist Number 1 Kims 2015: Bigger problem is tech adoption in healthcare.
Not tech applications or use cases,as far as I understand from my miniscule experience!


[9/8, 11:29 PM] Thanga Prabhu: It has crossed that bridge a long time ago. Even tier 2 & 3 towns have lab autoanalyzers and issue printed reports. In kollegal 80 km away from Mysore they have 5G connectivity. Right next to sathyamangalam forest. TeleICU is delivered connecting Mysore to Kollegal. Let us now celebrate the tech adoption Indian ecosystem has achieved without too much fuss. Building on top of it is the opportunity today.


[9/8, 11:31 PM] Metapsychist Number 1 Kims 2015: Agree,time to celebrate!!🥳🥳


[9/8, 11:36 PM] Metapsychist Number 1

 Kims 2015: What does 'selection bias' mean in research methodology?
Need it for some other work


[9/9, 5:52 AM] Thanga Prabhu: Bread and butter of statistics. Any textbook will have it...


[9/9, 6:46 AM] SBB: Selection bias in statistics refers to a situation where the sample used in a study or analysis is not representative of the larger population it is meant to represent. This bias occurs when certain groups or individuals are more likely to be included or excluded from the sample, leading to inaccurate or skewed results. It can compromise the validity of statistical inferences and make it difficult to generalize findings to the broader population. Researchers often use various techniques to minimize selection bias and ensure that their samples are more representative.


[9/9, 6:46 AM] SBB: 👆ChatGPT answer


[9/9, 8:14 AM] Rakesh Biswas: Next prompt : @⁨Metapsychist Number 1 Kims 2015⁩ 

What are the techniques to minimize selection bias?


[9/9, 8:15 AM] Rakesh Biswas: 👆Another real world evidence for selection bias in academic intervention trials!


[9/9, 8:15 AM] Rakesh Biswas: Selection bias driven by conflict of interest


[9/9, 8:17 AM] Rakesh Biswas: Factors influencing outcomes of academic intervention trials


[9/9, 8:17 AM] Metapsychist Number 1 Kims 2015: Minimizing selection bias is crucial when conducting research or making generalizations. Here are some general strategies to minimize selection bias:

1. **Random Sampling:** Use random sampling techniques to select your sample from the population of interest. This ensures that each member of the population has an equal chance of being included in the sample, reducing the risk of bias.

2. **Stratified Sampling:** If your population has subgroups with distinct characteristics, use stratified sampling to ensure representation from each subgroup. This can help prevent bias that might result from one subgroup being overrepresented or underrepresented.

3. **Use Random Assignment:** In experimental studies, randomly assign participants to treatment and control groups. This helps ensure that any differences in outcomes are not due to pre-existing differences between the groups.

4. **Increase Sample Size:** A larger sample size can help reduce bias. With a larger sample, the chances of obtaining a more representative and less biased sample increase.

5. **Blind and Double-Blind Studies:** In experiments, use blind or double-blind techniques to prevent participants or researchers from knowing which group they belong to. This minimizes the risk of bias in reporting or interpreting results.

6. **Avoid Convenience Sampling:** Avoid selecting participants or data points solely based on convenience or accessibility. Convenience sampling can introduce significant bias because it doesn't represent the entire population.

7. **Longitudinal Studies:** If possible, conduct longitudinal studies that track changes over time. This can help account for potential bias introduced by variations at a single point in time.

8. **Transparent Methodology:** Clearly document your sampling methods and research design to make it transparent and replicable. This allows others to assess the potential for bias in your study.

9. **Consider Nonresponse Bias:** Be aware of nonresponse bias if some participants or data points do not respond or drop out of your study. Analyze the characteristics of nonrespondents to assess potential bias.

10. **Validate Data:** Cross-check data through multiple sources or methods to ensure consistency and accuracy, reducing the likelihood of bias due to data errors.

11. **Peer Review:** Have your study or research design reviewed by peers or experts in the field. They can provide valuable feedback and help identify potential sources of bias.

12. **Sensitivity Analysis:** Conduct sensitivity analysis to assess how sensitive your results are to different assumptions and potential sources of bias.

By implementing these strategies, you can minimize selection bias and improve the validity and reliability of your research or analysis.

[9

[9/9, 8:34 AM] Rakesh Biswas: 👏👏full marks to chgpt

Now extrapolating to our current day academic intervention trials :

1) Randomly select all NEET PG aspirants 

2) Stratified selection based on special characteristics: general, reserved, EWS etc 

3) Random allocation to various departments! 

4) Devi Shetty? 

5) Don't let them know how they got selected into your academic department 

6) Stop this capitation business of convenience 

7) PaJR ProJR group and online learning portfolios for all students toward their life long CME monitoring 

8) See 7

9) Follow the drop outs who join IAS or become coaching center teachers instead of the noble prof 

10) Triangulation 360

11) See 10

12) Next prompt to cgpt : What's sensitivity analysis? 

The last paragraph by chgpt is a trashy opinion without evidence!



The image above is an   "illustration of hypothetical damage pattern on a WW2 bomber. Based on a not-illustrated report by Abraham Wald (1943), picture concept by Cameron Moll (2005, claimed on Twitter and credited by Mother Jones), new version by McGeddon based on a Lockheed PV-1 Ventura drawing (2016), vector file by Martin Grandjean (2021). Made available under creative commons licence here: https://commons.m.wikimedia.org/wiki/File:Survivorship-bias.svg#mw-jump-to-license

More about the image quoted from Wikipedia:

"During World War II, the statistician Abraham Wald took survivorship bias into his calculations when considering how to minimize bomber losses to enemy fire.[17] The Statistical Research Group (SRG) at Columbia University, which Wald was a part of, examined the damage done to aircraft that had returned from missions and recommended adding armor to the areas that showed the least damage.[18][19][20] The bullet holes in the returning aircraft represented areas where a bomber could take damage and still fly well enough to return safely to base. Therefore, Wald proposed that the Navy reinforce areas where the returning aircraft were unscathed,[17]: 88  inferring that planes hit in those areas were the ones most likely to be lost. His work is considered seminal in the then-nascent discipline of operational research."

Analogy to medical education and other life  
interventions: Intervention outcomes as to 
why mortality or drop out can be 
gleaned from "till end of trial participant 
survivors" and postulated as to 
why some didn't survive the system from the 
tell tale  markers of injury in those who didn't
drop out! 




No comments:

Post a Comment