02-14-2017, 08:17 PM
This highly regarded linguist analyzed Patsy, JMK, Nancy aka Norma Files and the Patricia Letters against the ransom note. Her findings are scientific, not opinion based since they are put through a computer.
The Patricia Letters were included since both Lou Smit and the Ramsey's believed the author could be the killer of JonBenet. These were various letters written to close case followers, media and investigators by a person pretending to be Patsy.
JMK and Norma Files were included in this comparison because both of them professed to have written the ransom note.
Patsy insists she did NOT write the note yet some in BPD believed she was the author.
The findings were that Both Patsy and Norma Files could be completely eliminated.
The findings were that both JMK and the author of the Patricia Letters were one in the same
The findings were that JMK was that JMK was the author of the ransom note.
Carole Chaski:
Given the candidate pool of Patsy Ramsey, Nancy and John Mark Karr documents
prior to 2000, (including the Tracey emails), the statistical procedure
using discriminant function analysis with leave- one-out cross-validation)
obtains a model with 100% accuracy. This means that each Patsy document was
tested as an unknown and returned to the Patsy class; each Nancy document
was tested as an unknown and returned to the Nancy class, and each John Mark
Karr document was tested as an unknown and returned to the John Mark Karr
class. This statistical model, at 100% cross-validated accuracy, is
therefore very good at recognizing Patsy's, Nancy's and Karr's documents.
This statistical model is then used to classify the unknown document, the
ransom note. The ransom note is classified as John Mark Karr's.
Given the candidate pool of Patsy Ramsey, Patricia and John Mark Karr
documents prior to 2000, including the Tracey emails), the statistical
procedure obtains a model of 58.3% accuracy. This model is not accurate
enough to make a predictive classification (it is just a bit higher than
chance). In my experience in validation testing of this authorship
identification method, extremely low model accuracy usually indicates that
there is a mixture in the data. Therefore, I combined the Patricia and John
Mark Karr documents prior to 2000, including the Tracey emails.
Given the candidate pool of Patsy Ramsey and the combined dataset of
Patricia and John Mark Karr documents prior to 2000, including the Tracey
emails), the statistical procedure obtains a model of 100% accuracy. Out of
these two potential authors, the ransom note is classified as Patricia/John
Mark Karr's.
I then tested the Patricia/John Mark Karr excluding the Tracey emails
document against the Patsy Ramsey documents. Again, the statistical model
obtained 100% cross-validated accuracy, and the ransom note is classified as
Patricia/John Mark Karr's.
I conducted 16 tests, most using pairwise data, in total, but these are the
most important results. It is my opinion that:
John Mark Karr and Patricia are one author, Nancy can at this time and in
this author pool be excluded as the author of the ransom note, Patsy Ramsey
can at this time and in this author pool be excluded as the author of the
ransom note, and John Mark Karr cannot at this time and in this author pool
be excluded as the author of the ransom note.
The method cannot predict if the document was authored by an author outside
the author pool; the method can only attach the questioned document to one
of the authors in the candidate pool. This is the state of the art in
linguistic authorship identification and a limitation which no one has yet
solved (be very wary of claims otherwise). Therefore it is extremely
important that all possible candidates be included in the author pool, and
that testing be done on each one in relation to all the others.
The Patricia Letters were included since both Lou Smit and the Ramsey's believed the author could be the killer of JonBenet. These were various letters written to close case followers, media and investigators by a person pretending to be Patsy.
JMK and Norma Files were included in this comparison because both of them professed to have written the ransom note.
Patsy insists she did NOT write the note yet some in BPD believed she was the author.
The findings were that Both Patsy and Norma Files could be completely eliminated.
The findings were that both JMK and the author of the Patricia Letters were one in the same
The findings were that JMK was that JMK was the author of the ransom note.
Carole Chaski:
Given the candidate pool of Patsy Ramsey, Nancy and John Mark Karr documents
prior to 2000, (including the Tracey emails), the statistical procedure
using discriminant function analysis with leave- one-out cross-validation)
obtains a model with 100% accuracy. This means that each Patsy document was
tested as an unknown and returned to the Patsy class; each Nancy document
was tested as an unknown and returned to the Nancy class, and each John Mark
Karr document was tested as an unknown and returned to the John Mark Karr
class. This statistical model, at 100% cross-validated accuracy, is
therefore very good at recognizing Patsy's, Nancy's and Karr's documents.
This statistical model is then used to classify the unknown document, the
ransom note. The ransom note is classified as John Mark Karr's.
Given the candidate pool of Patsy Ramsey, Patricia and John Mark Karr
documents prior to 2000, including the Tracey emails), the statistical
procedure obtains a model of 58.3% accuracy. This model is not accurate
enough to make a predictive classification (it is just a bit higher than
chance). In my experience in validation testing of this authorship
identification method, extremely low model accuracy usually indicates that
there is a mixture in the data. Therefore, I combined the Patricia and John
Mark Karr documents prior to 2000, including the Tracey emails.
Given the candidate pool of Patsy Ramsey and the combined dataset of
Patricia and John Mark Karr documents prior to 2000, including the Tracey
emails), the statistical procedure obtains a model of 100% accuracy. Out of
these two potential authors, the ransom note is classified as Patricia/John
Mark Karr's.
I then tested the Patricia/John Mark Karr excluding the Tracey emails
document against the Patsy Ramsey documents. Again, the statistical model
obtained 100% cross-validated accuracy, and the ransom note is classified as
Patricia/John Mark Karr's.
I conducted 16 tests, most using pairwise data, in total, but these are the
most important results. It is my opinion that:
John Mark Karr and Patricia are one author, Nancy can at this time and in
this author pool be excluded as the author of the ransom note, Patsy Ramsey
can at this time and in this author pool be excluded as the author of the
ransom note, and John Mark Karr cannot at this time and in this author pool
be excluded as the author of the ransom note.
The method cannot predict if the document was authored by an author outside
the author pool; the method can only attach the questioned document to one
of the authors in the candidate pool. This is the state of the art in
linguistic authorship identification and a limitation which no one has yet
solved (be very wary of claims otherwise). Therefore it is extremely
important that all possible candidates be included in the author pool, and
that testing be done on each one in relation to all the others.