Simpson’s paradox

-

 

Simpson’s paradox



An article in the British Medical Journal [“Comparison of treatment of renal calculi by operative surgery, percutaneous nephrolithotomy, and extracorporeal shock wave lithotripsy” (1986, Vol. 82, pp. 879–892)] provided the following discussion of success rates in kidney stone removals. Open surgery had a success rate of 78% (273/350) and a newer method, percutaneous nephrolithotomy (PN), had a success rate of 83% (289/350). This newer method looked better, but the results changed when stone diameter was considered. For stones with diameters less than 2 centimeters, 93% (81/87) of cases of open surgery were successful compared with only 83% (234/270) of cases of PN. For stones greater than or equal to 2 centimeters, the success rates were 73% (192/263) and 69% (55/80) for open surgery and PN, respectively. Open surgery is better for both stone sizes, but less successful in total. In 1951, E. H. Simpson pointed out this apparent contradiction (known as Simpson’s paradox), and the hazard still persists today. Explain how open surgery can be better for both stone sizes but worse in total.

EXPLANATION:

success rate

 

STONE dia < 2

STONE dia>=2

TOTAL

Open Surgery

93% (81/87)

73% (192/263)

78% (273/350)

Newer PN Method

83% (234/270)

69% (55/80)

83% (289/350)

 

 

 

 

 

Here, Open Surgery is better for both cases. Total is higher is lesser as total samples for Stone dia>2 is more ie 263 comparing to Stone dia<2 ie 87. So, the Total will be nearer to 73 and far from 93 ie 78. In Newer PN Method since total samples for Stone dia<2 is more ie 270 than Stone dia >2 ie 80, the Total is more nearer to 83 than 69. Hence, the total depends on no of samples in each group. Since, It is not equally grouped, the total is unreliable.

Technical Explanation:

According to Multiplication rule & Total Probability Theory, The overall success rate depends on the success rates for each stone size group, but also the probability of the groups.It is the weighted average of the group success rate weighted by the group size as follows P(overall success) = P(succes rate of stone dia>=2)*P(stone dia>=2))+ P(success rate of stone dia < 2)*P(stone dia < 2 ). 

Here, Probability of sample percentage of each stone size in each method varies.

For open surgery, the dominant group (stone dia>=2) has a smaller success rate while for PN, the dominant group (stone dia < 2) has a larger success rate.


Location: Amritanagar, Ettimadai, Tamil Nadu 641112, India

Comments

Post a Comment

Popular posts from this blog

Some Logic Puzzles

-
Image
Hi folks!!  Here are some Logic puzzles to solve and sharpen your mind. All of these needs only basic Computer and Discrete Logics to solve and It helps you in improving your Logical Reasoning, Aptitude, Algorithmic thinking and problem-solving skills. It is ordered according to levels so that reader can approach in order. I have solved few questions and remaining similar questions are left to readers to try. Those are common and famous puzzles solutions and explanations can easily be found in web and books. LEVEL-1 Each inhabitant of a remote village always tells the truth or always lies. A villager will give only a “Yes” or a “No” response to a question a tourist asks. Suppose you are a tourist visiting this area and come to a fork in the road. One branch leads to the ruins you want to visit; the other branch leads deep into the jungle. A villager is standing at the fork in the road. What one question can you ask the villager to determine which branch to take? The question is “If I
Read more

Data Preprocessing

-
Image
  DATA PREPROCESSING The objective of data pre-processing is to analyze, filter, transform, and encode data so that a machine learning algorithm can understand and work with the processed output. The phrase "garbage in, garbage out" is very much apt to data mining and machine learning projects. The presence of any unclean data like missing attributes, attribute values, containing noise or outliers, and duplicate or wrong data will degrade the quality of the ML results. So, It is important to manipulate or transform the raw data in a useful and efficient format before it is used in Machine learning to ensure or enhance performance. Important Libraries for Data Preprocessing: To do data preprocessing in Python, we need to import some predefined Python libraries. These libraries are used to perform some specific tasks. There are three specific libraries that we will use for data preprocessing. ·        Numpy : The Numpy Python library is used to include many kinds of mat
Read more

Artificial Intelligence - An Overview by Dharaneish

-
Image
Hi all!! Here is a documentary-cum-newsletter done on AI by me. Artificial Intelligence - An Overview by  Dharaneish V C . It is about an basic overview about fundamentals, history, types, understanding of AI, ML, DL and it's applications, advantages, disadvantages, challenges and economic impact. It is intended for any beginner, high school student, UG from any field, literally anyone with no pre-requiste!!! Hope it will be useful to you. Feel free to add your sugesstions in comments.  
Read more