Editing 2893: Sphere Tastiness

This comic graphs the tastiness vs. the size of four roughly spherical objects: {{w|melons}}, {{w|grapes}}, {{w|Earth|Earth}} and the {{w|Moon}}. Based on the the fact that melons and grapes are (in this context) relatively small and tasty to most people, and that planetary scale bodies are relatively large and made mostly of rocks and metals generally considered not remotely tasty,{{cn}} [[Randall]] postulates the existence of an intermediate body, one which is approximately 800 meters in diameter and "tastes okay".

This is the second comic in a row to feature fruit, graphs and predictions (after [[2892: Banana Prices]]), and continues the theme of a logarithmic axis scale to facilitate plotting a linear regression. Here the line is interpolated between known data, rather than extrapolated beyond it. Such interpolation is quite common in scientific analysis, and is often useful, but this example clearly leads to a ludicrous conclusion. Using such ridiculous analyses to show the dangers of flawed and/or sloppy methodology is a common theme in xkcd.

* there are only four data points, which is insufficient to interpolate from.

* these clusters represent entirely different sub-classes of spherical object (fruit vs. astronomical bodies) while other subclasses are not represented at all (the title text mentions this flaw).

* as tight clusters of [[2533: Slope Hypothesis Testing|similarly sourced data]], it effectively reduces the data down to two useful data points. This also makes the choice of log-median interpolation unjustified.

* the 'tastiness' scale has no indication of what assessment (subjective or objective) it records. Nor does it even have graduations, making it unknown if the graph is linear-log or log-log (or otherwise), changing the implied meaning behind the choice of straight-line interpolation.

* according to astronaut John Young, who visited the Moon's surface during the Apollo 16 mission, [https://phys.org/news/2006-02-mysterious-moondust.html "moondust doesn't taste half bad"]. (Although other Apollo astronauts likened its smell and taste to burnt gunpowder, so make of that what you will.)

The title text points out that {{w|baseball (ball)|baseballs}} seem to refute this theory since they're not usually thought of as tasty, but they're between the sizes of grapes and melons, which would place them in the bottom left of the graph, way off the fit line. Baseballs are typically made of a combination of a rubber or cork center wrapped in yarn, and covered by either horsehide, cowhide or synthetic leather. In point of fact, there are many, many common round objects that completely fail to conform to this graph, but rather than acknowledge that this analysis is fatally flawed, Randall uses {{w|special pleading}} to justify its exclusion from the graph, suggesting that the problem is that we lack "the right seasonings". While seasonings can improve the taste of foods, it's implausible that the inedible components of baseballs would be rendered "tasty" with any conceivable combination of seasonings. Even if they could, there's no evidence that such would give them the proper level of 'tastiness' to conform to the graph. This argument lampoons the use of "cherry picking" and motivated reasoning, in which researchers include only data points which fit their hypothesis and make up reasons to exclude those which don't. This is obviously very poor science, but less exaggerated versions are all too common in scientific studies.  

The comic refers to this plot as research. This is an exaggeration, since two clusters of paired points are rarely considered sufficient for research purposes. But plotting a justifiably sufficient quantity of data points on a logarithmic plot, and then drawing a line through them, is a common way to visualize an actual exponential relationship more comprehensibly. An example of that is the {{w|Gutenberg–Richter law}} where the magnitude of earthquakes (an intrinsically logarithmic scale) in a particular region is plotted together with the frequency of occurrence, typically resulting in a statistically significant straight line.

Other fruit opinions have previously been mentioned in [[388: Fuck Grapefruit]], but it is unknown what the line would be like if Randall included grapefruit.

Other absurd uses of linear regression are seen in [[605: Extrapolating]] and [[1204: Detail]].

:[Graph with Y axis using an arrow indicating tastiness from "Not Tasty" to "Tasty" and X axis labeled "Sphere Diameter (meters)" with a logarithmic scale running from 10^-5 to around 10⁸ (with 10^-3, 10⁰, 10³ and 10⁶ labeled).]