WHAT IS AN IMPUTATION?...
....Professor Abbott in his article published in a US IEEE online magazine, made great stock of the fact that he had arrived at the identity of the Somerton Man using a single, rootless, hair shaft that was extracted from the plaster bust of the body supposedly that of the Somerton Man. the bust was a slip cast and was made by the late Mr. Paul Lawson. The professor also announced in this article that he had used imputation to arrive at his result....
LINK TO ARTICLE:
https://spectrum.ieee.org/somerton-man
https://spectrum.ieee.org/somerton-man
The diagram above nicely describes the process of imputation.
You start with some incomplete data, you add an educated guess or two, produce a creative result and make your claim. In other words, it could be wrong.
Imputation, in its simplest form, is educated guesswork. It's a statistical technique that can be used to fill in the missing values in a dataset.
Let's say that you're conducting a survey, and some people didn't answer every question. Instead of throwing away their entire response or ignoring the unanswered question, you can use imputation to make an educated guess about what their answer might have been based on the other information you have about that person.
Let's say that you're conducting a survey, and some people didn't answer every question. Instead of throwing away their entire response or ignoring the unanswered question, you can use imputation to make an educated guess about what their answer might have been based on the other information you have about that person.
Would you consider that to be accurate? It depends on a few factors:
The quality of the available information: If the data is rich and well-structured, the imputation will likely be more accurate, not perfect but possibly more accurate
The outcome also depends on the imputation method: There are simple methods, like filling in the missing value with the mean (average) or median of the other values for that variable, which are less accurate. Then, there are more complex methods, like regression imputation or multiple imputation, which consider other variables in the dataset to make a more informed guess and tend to be more accurate.
The amount and pattern of missing data: If you have very little missing data, imputation might not affect your results much. But if a lot of data is missing or if the missingness is not random (like if certain types of people were more likely to skip a question), imputation could be less accurate and might introduce bias.
In this case, there are potential variables at every turn. The possibility of human hair being part of the slip-cast mix is a major issue. Another issue is what appears to be the closure of the company that performed the analysis of the rootless hair shaft. Bear in mind that extracting useful DNA from a rootless hair shaft is still considered with skepticism by many in the forensics field.
In summary, imputation isn't a perfect solution. It makes assumptions and adds artificial data to your dataset. While it can improve the analysis by providing a complete dataset, it can never fully recapture the true information that would have been provided by the original, non-missing data. Hence, it's critical to consider the potential implications and limitations when using imputation methods.
I will close this post off with the words of a Forensic Scientist, Dr. Xanthe Mallet, who made a point of saying that the DNA results as they stand are not conclusive you'll be able to see and hear what she has to say at the 18 minutes 52 seconds mark. The documentary is online here:
'MY CONCERN IS THAT WE MAY NEVER BE ABLE TO CATEGORICALLY SAY THAT WE KNOW THIS PERSON'S IDENTITY...'
Dr.Xanthe Mallett
With thanks to Peter Davidson for his valuable input...