Attempts to use it in an expression would always throw an exception.
I think you've misunderstood my post. If I wanted to causes exceptions, I'd leave the zeros in place. or replace them with #NaN.
My point was that by replacing the zeros with a very, very small number, you can avoid exceptions, or the need for conditional tests to avoid exceptions, as it will allow most calculations to operate correctly whilst introducing negligible changes to the final results.
With the rise and rise of 'Social' network sites: 'Computers are making people easier to use everyday'
Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.
"Science is about questioning the status quo. Questioning authority".
In the absence of evidence, opinion is indistinguishable from prejudice.
| [reply] |
I did misunderstand your intention. However I still do not understand your solution. If I assume that the zero data is really a small positive number that displays as '0.00', its log should be a negative number whose magnatude exceeds any of the other logs. (e.g. log(0.001) = -6.9) Setting the log to zero assumes that the raw value is e (2.71...). Note: All other values are between .001 (log(.01) = -4.6) and .99 (log(.99) = -.01).
| [reply] |
Setting the log to zero ...
I didn't suggest setting the log to zero. Or mention doing anything to the logs.
its log should be a negative number whose magnatude exceeds any of the other logs.
And if you set values that are 0(.00000) to 1e-308, then when their logs are taken, they become
say log(1e-308);;
-709.196208642166
which fulfills your "negative number whose magnatude exceeds any of the other logs." in spades.
With the rise and rise of 'Social' network sites: 'Computers are making people easier to use everyday'
Examine what is said, not who speaks -- Silence betokens consent -- Love the truth but pardon error.
"Science is about questioning the status quo. Questioning authority".
In the absence of evidence, opinion is indistinguishable from prejudice.
| [reply] [d/l] |
HUGE_VAL does not help here, we need kind of a TINY_VAL. 1 / HUGE_VAL does not work:
$ perl -MPOSIX=HUGE_VAL -e 'print log(1 / HUGE_VAL)'
Can't take log of 0 at -e line 1.
| [reply] [d/l] [select] |