-hh
Contributor
- Messages
- 1,021
- Reaction score
- 254
Apologies for the delay in responding; was on the road with business travel.
Understood, but you missed my point: the data analysis I'm interested in requires additional data that DAN didn't publish (and likely didn't collect). So its not merely a question of re-hashing the data provided to put it into a different format.
For example, my one comment was that 'Expert' divers were observed to have had a higher absolute number of accidents reported, but because we don't know how large of a subset they are, we can conclude nothing about their incidence rates.
What's missing from the DAN data set in this case are the total number of dives from which some accidents occurred, and what percentage of that total were performed by each of the identified subsets that DAN made.
Similarly, when you look at this same subset element over time, DAN switched measures from #dives to Certification Level. Perhaps you're comfortable with those two metrics being objectively identical, but I am not, so I'm not going to claim that today's report has resulted in a 'change' in risks from years ago, as opposed to the difference being due (at least in part) to differences in the reported subsets.
Same here.
Bottom line to all of this is that no amount of re-hashing of data can ever fill in the blanks of unpublished/uncollected data.
-hh
The DAN report describes every one of the deaths. No need to be misled by any sampling numbers. All it takes is the time to go through them to see what's what. You can make your own tables if you like to show whatever trends you see there. If you read through this entire thread, you will see that was discussed earlier.
Understood, but you missed my point: the data analysis I'm interested in requires additional data that DAN didn't publish (and likely didn't collect). So its not merely a question of re-hashing the data provided to put it into a different format.
For example, my one comment was that 'Expert' divers were observed to have had a higher absolute number of accidents reported, but because we don't know how large of a subset they are, we can conclude nothing about their incidence rates.
What's missing from the DAN data set in this case are the total number of dives from which some accidents occurred, and what percentage of that total were performed by each of the identified subsets that DAN made.
Similarly, when you look at this same subset element over time, DAN switched measures from #dives to Certification Level. Perhaps you're comfortable with those two metrics being objectively identical, but I am not, so I'm not going to claim that today's report has resulted in a 'change' in risks from years ago, as opposed to the difference being due (at least in part) to differences in the reported subsets.
...Research has been an important part of much of my life. It has at times been a part of my job description, as it is now...
Same here.
Bottom line to all of this is that no amount of re-hashing of data can ever fill in the blanks of unpublished/uncollected data.
-hh