Saturday, December 24, 2016

Numerical Analysis and Error for the Dummy Statistician

numerical analysis
Numerical analysis
Wow: a week of Techwalla silliness and we've only barely scratched the surface... maybe we should have made it Techwalla month? Whatever the case, number seven in the series is courtesy of seven-time awardee Tom Lutzenberger, now appearing in a sixth category! This time it's mathematics, a topic we rather suspect the PolySci-slash-English major with the MBA avoided like the plague after finishing high school... and that lack of knowledge was pretty obvious when he posted "Types of Errors in Numerical Analysis" at Techwalla.com.

Lutzenberger gets right to the point in his introduction, explaining that
"In the world of math, numerical analysis is well known for focusing on the algorithms used to solve issues in continuous math."
Although one of our staff does have a single-digit Erdös number (7, by way of Walter Alvarez [5] and Ron Blakey [6]), even our non-mathematicians recognize that sentence for the bull it is. Tom doesn't get to the concept of approximation for several paragraphs, and clearly has no idea what "continuous math" could possibly mean. Given this lack, Tom simply went straight to Wikipedia and attempted to reword it. You can tell, because his topics are in the same order and he even attempts to use one of the examples! So let's see what Mr. Lutzenberger has to say:
"The round-off error is used because representing every number as a real number isn't possible. So rounding is introduced to adjust for this situation. "
     Here, Tom's attempting to reword the wikipedia entry, which says that "Round-off errors arise because it is impossible to represent all real numbers exactly on a machine with finite memory" -- not that one "uses" a rouding error, but that it's "introduced"! Next, Lutzenberger takes on truncation, claiming that
"A truncation error occurs when approximation is involved in numerical analysis. The error factor is related to how much the approximate value is at variance from the actual value in a formula or math result."
But wait: isn't the whole idea of numerical analysis approximation? And that's not what truncation is, anyway -- truncation errors are introduced when a series of calculations is truncated, not a number. Here's where Lutzenberger attempts to use an example, claiming that if you...
"...take the formula of 3 x 3 + 4. The calculation equals 28. Now, break it down and the root is close to 1.99. The truncation error value is therefore equal to 0.01..."
Well, no: for one, the formula in the wikipedia example isn't "3 x 3 + 4" and it isn't "3 times 3 plus 4" as appeared in the original at eHow.com (thanks to archive.org): the example, which is still on wikipedia, shows the process of numerical analysis being applied to the formula 3x³+4 = 28! What an idiot! There's more (of course), but we're pretty sure you get the idea...

[Note: apparently the content editors at Techwalla or eHow.com decided to "spiff up" Lutzenberger's prose some time in 2015; introducing even more bullshit to what was already rubbish]

Worst of all, Lutzenberger got paid to misinform his readers. While he may have appreciated the bump to his PayPal account, we here at the Antisocial Network do not in the least appreciate his contribution to the stupidification of the internet -- and that's why Tom Lutzenberger is the (not-so-) proud recipient of yet another Dumbass of the Day award.     
copyright © 2016-2021 scmrak

MM - STATISTICS

No comments: