[airza]

Why would the median temperature be a useful measurement here?

Quick, what is the median of these numbers: -2,-1,0,1,2 And what is the median of these numbers: -7,-1,0,1,2

Of course the average and median are identical for a perfectly distributed normal distribution. But part of science is supposed to be reducing the noise from your signal. Why are you adding noise in your quest to investigate the average temperature of the earth?

[glitchc]

In the presence of outliers the median is a better predictor of underlying long-teem trends. Subtracting the median gives the delta from zero-median. The delta would be non-zero if there is a long term effect. Here the delta is zero in a number of cases.

[airza]

What is the concept of an outlier when you’re trying to measure if the earth is hotter than it used to be over a fixed length of time? If your hottest temperature 500 degrees, then sure, you may have an outlier that needs exclusion. But if all your coldest days are the same and your hottest days are 10 degrees warmer, your median might not move much. But that doesn’t make it look good for life on earth.

[CrazyStat]

> In the presence of outliers the median is a better predictor of underlying long-teem trends.

Not if outliers are REALLY FUCKING IMPORTANT, like, say, extreme weather events.

[thehappypm]

Outliers are likely to drag the averages down, not up. The only really substantial global outlier events are volcanic eruptions, which cause a net cooling. Everything else (hurricanes and the like) is just heat moving around

[CrazyStat]

We're looking at individual airport temperature records here, not global means.

[thehappypm]

So? A big volcanic eruption could cool many airports.

[CrazyStat]

Once we recognize that we're talking about local rather than global there are many local outliers that increase temperatures, countering your argument that volcanic eruptions are the only global outliers.

[glitchc]

Extreme weather events are a symptom of global warming, not a cause.

[WalterGR]

Yes and?

[glitchc]

And that means their measurement contributes to the noise when attempting to identify the long-term causal trend. Taking the median is the correct step. One doesn’t measure the output and assume it’s the input.

[cpurdy]

Q: "Why would the median temperature be a useful measurement here?"

A: The median temperature is the calculation that allowed him to deny planetary warming while using statistics.

Look, when you want to show something to be true that is actually false, statistics will help you to do so.