In a previous
post I discussed my work on staying power models, and reported how for the two data sets I used the results were indecisive, mainly due to the small size of the data sets. The two data sets were for 1942 American and Japanese carriers subject to bomb attack and Royal Naval cruisers hit by torpedoes (which detonated).
I have now extended the cruiser torpedo hit data set from RN cruisers to now include USN, Japanese and Italian cruisers as well. This data set comprises about 90 records and now shows staying power more or less scaling as the ship displacement. Better yet I have analysed the posterior distribution of the scaling power (the power of displacement that staying power is proportional to) and show that with the previous cruiser data set the analysis was consistent with a wide range of scaling power (at least from 0.3 up to 0.9), while the posterior distribution with the extended data set peaks tightly between ~0.8 and 1.1.
The question now seems to be should I prepare what I now have for publication or continue extending the work.
 |
| Scatter Plot of the Cruiser Data and 25, 50 and 75% Probability of Sinking Contours |
 |
| Approximate Posterior Distribution of Staying Power RN Cruiser Data Only |
 |
| Approximate Posterior Distribution of Staying Power Full Cruiser Data Set |
