Paul -- A Basic Stylometric Study

[Peter Kirby]

At this point I think I'm interested in working on the program some more... I've seen enough to know that it has promise but could use work.

[Ben C. Smith]

In this thread, I have solely been making that call based on this number (e.g., this is a 'negative' result):

Bayesian Comparison of Best Author to Best Control: from Equal Priors, Z-Score-Based Method
$VAR1 = 2; $VAR2 = '0.33154046916538'; $VAR3 = 25; $VAR4 = '0.66845953083462';

Less than 0.5 is negative (the best "control" provides the closest match to the sample).

Greater than 0.5 is positive (the best "author candidate" provides the closest match to the sample).

So this result is negative because...?

1. $VAR1 came out as #2 (Acts, in this case), not #1 (Paul).
2. $VAR2 came out as less than 0.5.
3. Both A and B.

[Secret Alias]

The other possibility that you haven't considered is that the Pastorals and Ephesians were developed around a chunk or chunks of authentic Pauline material.

[Peter Kirby]

In this thread, I have solely been making that call based on this number (e.g., this is a 'negative' result):

Bayesian Comparison of Best Author to Best Control: from Equal Priors, Z-Score-Based Method
$VAR1 = 2; $VAR2 = '0.33154046916538'; $VAR3 = 25; $VAR4 = '0.66845953083462';

Less than 0.5 is negative (the best "control" provides the closest match to the sample).

Greater than 0.5 is positive (the best "author candidate" provides the closest match to the sample).

So this result is negative because...?

1. $VAR1 came out as #2 (Acts, in this case), not #1 (Paul).
2. $VAR2 came out as less than 0.5.
3. Both A and B.

The correct answer is B.

It's "negative" in the sense that we can't say that the closest 'author candidate' is any better than some random "control" (because the "control" is actually the closest match out of them all), whom we know is not the author.

The purpose of this is to have some way to tell when none of the 'author candidates' are very likely to be the true author, since this is what we usually want to know in attribution studies of ancient texts: is it this guy, that guy... or neither, source unknown?

This is actually considered a very hard problem in stylometry (being able to determine whether none of the somewhat-reasonable author candidates is the actual author), sometimes called the problem of "open" authorship attribution ("open" to the possibility that none on the candidate list are the actual author). I think the approach taken worked remarkably well, for how simple it is.

[Secret Alias]

I don't know why we continue to suppose that 'to the Corinthians' 'to the Romans' are units developed by Paul rather than a Catholic editor. Look at the falsification effort with respect to the Ignatian canon going from Syriac to Greek. New works get created - viz. to the Trallians. http://stephanhuller.blogspot.com/p/sho ... ho-is.html Irenaeus speaks of similar patterns with the gospels. The editor(s) took things out from their original context and moved them around like stones in a mosaic reused to make a whole new 'image.' The pattern in the falsification of early Christian material follows the cento poem model and Irenaeus was a master at 'centonizing' verse.

[Peter Kirby]

I don't know why we continue to suppose that 'to the Corinthians' 'to the Romans' are units developed by Paul rather than a Catholic editor. Look at the falsification effort with respect to the Ignatian canon going from Syriac to Greek. New works get created - viz. to the Trallians. http://stephanhuller.blogspot.com/p/sho ... ho-is.html Irenaeus speaks of similar patterns with the gospels. The editor(s) took things out from their original context and moved them around like stones in a mosaic reused to make a whole new 'image.' The pattern in the falsification of early Christian material follows the cento poem model and Irenaeus was a master at 'centonizing' verse.

(I know why, but I'm not telling. Hahahahaha.)

If you prefer, consider this more a study in how many different "hands" went into the authorship of these letters.

[Ben C. Smith]

The correct answer is B.

It's "negative" in the sense that we can't say that the closest 'author candidate' is any better than some random "control" (because the "control" is actually the closest match out of them all), whom we know is not the author.

The purpose of this is to have some way to tell when none of the 'author candidates' are very likely to be the true author, since this is what we usually want to know in attribution studies of ancient texts: is it this guy, that guy... or neither, source unknown?

This is actually considered a very hard problem in stylometry (being able to determine whether none of the somewhat-reasonable author candidates is the actual author), sometimes called the problem of "open" authorship attribution ("open" to the possibility that none on the candidate list are the actual author). I think the approach taken worked remarkably well, for how simple it is.

Got it. Thanks for the explanation(s).

[Peter Kirby]

The other possibility that you haven't considered is that the Pastorals and Ephesians were developed around a chunk or chunks of authentic Pauline material.

I considered that too. Maybe I haven't "mentioned," but I have considered.

[Tenorikuma]

A relevant citation would be J. C. O'Neill, "Paul Wrote Some of All but not All of Any", The Pauline Canon (Pauline Studies 1). If his theory (and those like his) of the formation of the Pauline corpus are correct, stylometry will probably be of little use.

Still, I'd like to see 2 Timothy and Titus compared to Acts. Edit: Doh! Sorry I missed that.

[Peter Kirby]

Still, I'd like to see 2 Timothy and Titus compared to Acts.

Oh, they were. No dice. Better match for Paul. Not-so-good compatibility with Acts.

Below:
$VAR1 = Paul (1 Cor, 2 Cor, Gal, 1 Thess)
$VAR2 = Hebrews
$VAR3 = Acts

But here's the combined 2 Timothy & Titus reading:

testsize: 1897


Bayesian Author Test: Posterior Probabilities from Equal Priors, Z-Score-Based Method
$VAR1 = '0.913206946618373'; $VAR2 = '0.000188298871011871'; $VAR3 = '0.0866047545106156';

Bayesian Comparison of Best Author to Best Control: from Equal Priors, Z-Score-Based Method
$VAR1 = 1; $VAR2 = '0.560961574800625'; $VAR3 = 25; $VAR4 = '0.439038425199375';

Percentage of Samples in the Best Author Candidate that Meet the P-Value>0.11 Test, Z-Score-Based Method
1
Percentage of Samples outside the Best Author Candidate that Meet the P-Value>0.11 Test, Z-Score-Based Method
0
Posterior Probability of a Sample Meeting the Test Being by the Best Author Candidate (with Prior = 0.5), Not Any Other, Z-Score-Based Method
1

Author Z-Score-Based P-Values
$VAR1 = '0.112024483130811'; $VAR2 = '2.30989084975012e-05'; $VAR3 = '0.0106239367721071';
Good compatibility. Z-Score-Based P-Value > 0.1.
Excellent indicator. 0% of the rest have P-Value > 0.1.

Control Z-Score-Based P-Values
$VAR1 = '0.00337474653502709'; $VAR2 = '0.0266200361752983'; $VAR3 = '9.0030172758527e-06'; $VAR4 = '0.0123644990002458'; $VAR5 = '0.00132850753017783'; $VAR6 = '0.00709492638297052'; $VAR7 = '0.00175257570262458'; $VAR8 = '0.00533457909709268'; $VAR9 = '0.00671201236496284'; $VAR10 = '0.00496067849716306'; $VAR11 = '3.49936936778673e-05'; $VAR12 = '0.00135314515113046'; $VAR13 = '4.07477857443668e-05'; $VAR14 = '9.70034537998303e-05'; $VAR15 = '0.00107058572184818'; $VAR16 = '0.0104469193460479'; $VAR17 = '0.00249362111301281'; $VAR18 = '0.00349743508675774'; $VAR19 = '0.0114736805833751'; $VAR20 = '2.37604830766775e-05'; $VAR21 = '0.0635974725031043'; $VAR22 = '0.0144236194199457'; $VAR23 = '0.0464990882304803'; $VAR24 = '0.0535010392360603'; $VAR25 = '0.0876763309055633'; $VAR26 = '0.0582419082013158'; $VAR27 = '0.0208341493994333'; $VAR28 = '0.0513936936733187'; $VAR29 = '0.00262889682683281'; $VAR30 = '0.00632074030281559'; $VAR31 = '0.00317891769079122'; $VAR32 = '0.00584224377384102'; $VAR33 = '3.98403945841793e-99'; $VAR34 = '0.000128983177380328'; $VAR35 = '1.95454177103593e-101'; $VAR36 = '0.0509308342409494'; $VAR37 = '0.0267388094870112'; $VAR38 = '0.0678909578596876'; $VAR39 = '0.037435219364268';

In all the tests of the "13 letters," none of them identified as closest to Hebrews or closest to Acts, even though they were in all the tests (and even though Acts certainly has enough material to provide a basis of comparison--Hebrews a little less so).