[CCR PATCH] Calculate appropriate po2 values

[Willem Ferguson]

Robert,

On 21/06/2014 12:43, Robert C. Helling wrote:
>
> three quick comments:
>
> 1) What about the case of readings 1050, 1140, 1160? Then only one difference is big.
As is, diff_limit is the value that determines the way the rule is 
applied. I hope that diff_limit will ultimately be an editible value in 
the Subsurface Preferences tab, but with a sensible default value. 
Assume here diff_limit is 100 mBar which, from my interacting with CCR 
logs, appears in the right ball park.

Your example: This is case A) of the rule. The mean of the three values 
would be calculated. [This is not the way the APD Discovery does it, the 
APD equipment would reject the 1050 value as an outlier, but on this 
discussion forum there has been some criticism of the simplistic 
rejection of outlier values implemented by APD]. The present code would 
treat 1050,1140,1160 in the same way as 1060,1140,1160.

The values 1030,1140,1160 would trigger case B) of the rule. Now the 
difference between the 1030 and the other two values are too big and the 
lower value is rejected as an outlier.

The values 1030,1140,1270 would trigger case C) of the rule, which is 
calculating the mean of the three values (in other words, actually, 
handled the same way as case A). In case C) there is no way of knowing 
which values is are outliers.

Any comments or suggestions would be very valuable here.

> 2) The logic of hard coding all the cases with explicit 1,2,3 is in my point of view not the most straight forward. Why don’t you start by sorting the three values (can be done in a single for loop: Find both the min and the max of the three and the middle one is the sum - min - max) and then have the clear cases min far away from mid or max far away from mid.
It seems a sensible way to go, except that the code currently does not 
use the middle reading as a reference (there is nor reason to believe 
that the middle value has any special significance). Rather, it uses the 
relative differences among all three of the readings. But the mechanism 
of using a do-loop is attractive.

Comments, please!

Kind regards,
willem