| I'm no engineer, but I'll take a crack at it: * Both axes of the chart are inaccurate. The temperature on the X axis intermingles ambient and O-ring temperature, but they're not the same (imagine ocean temperature vs air temperature). The vertical axis, O-ring damage, is semi-relevant, but the important question is whether the O-rings held a seal (degree of blow-by). O-ring damage probably correlates, but seems like a proxy that may be misleading. * The data available in the chart was not available to the engineers at the time. There was only one previous failure (SRM 15) that led anyone to believe temperature itself was an issue. They had two valid temperature data points (SRM 15 and SRM 22), and correctly pointed out that another launch at 29 degrees was completely outside their tested range and would be inherently risky. It's not clear to me that a scatterplot with two datapoints would be that effective. Furthermore Boisjoly criticizes that Tufte presumed to judge the engineers, for the following reasons: * The engineers previously recommended no launch should occur, months in advance (due to previous O-ring issues), but NASA overruled them. * The engineers recommended, and their managers accepted, that the low launch temperature was outside their test database. NASA came back and requested proof that the temperature was dangerous, but the engineers could not comply, because the parameters were extreme enough that they had not tested them. They couldn't prove a negative. The managers at Morton Thiokol and NASA then jointly overruled the engineers. * The idea of putting together a chart was not something the engineers considered, because there were a variety of previous unrelated problems in O-rings (each resolved afterwards) that muddied that data, and muddies the data that Tufte presents. * Tufte himself failed to research or note the pivotal information, and thus misrepresented the situation the engineers were in, and the data available, while imputing that the engineers should be held morally responsible for not presenting data they didn't have (and then going on to present such data himself, compiled after the fact and at his own leisure, incorrectly). |