Is (or was) there a major bug in Calc math?

[AHoerner]

I am still using LO 6.1.4.2.
.

number…log…log10

3100…3.49136169383427 3.49136169383427

3200…3.50514997831991 3.50514997831991

0.031746031746032 0.013788284485634 0.013788284485633

The top two numbers in the left hand col are just numbers. The numbers to the right of them are the log()s and log10()s, respectively, as produced by calc. The two right-hand numbers in the bottom row are the differences, also from calc.

The right-hand number in the bottom row is also from calc by the formula (3200-3100)/((3200+3100)/2), i.e., the difference as a percentage of the average value. It is correct. It is also the correct answer to the two differences to its right. That is what R says, to 4 significant figures, and also, the three numbers should be the same. And the exp( ) of the difference should give you the ratio. calc gets the ratio right - it is 1 + the percentage difference, roughly. But exp( ) of the 0.128 number is way wrong.

Is this a defect peculiar to my machine or installation, or was it in the calc itself? If the latter, is it still there?

[ve3oat]

And the exp( ) of the difference should give you the ratio.

Well, the antilog of the difference of two logs will give you the ratio. Is that what you meant?

But where does the quoted “0.128” come from?

[mikekaganski]

And better attach an ODS with data, instead of writing some numbers with explanations that could be difficult to understand, and to reproduce; then your task would be to explain in which cells the results are wrong in your opinion, and why.

[AHoerner]

This is my mistake. I did not realize that the equality between the difference in logs and the percentage change for small changes only held true for natural logs. I thought it was true of all logs, regardless of base. With base 10 logs, you need to multiply the difference by ln(10) to get the correct percentage. I had the false belief that you did this on both the top and bottom, canceling out.