Are there graphs with the median temperature development in the 20th century and if not, is there readily accessible data that can be used to calculate it?

Median is statistically more stable, at the same time I'd be curious to plot the variance as well. My suspicion is the median rises less steep than the average but at the same time the variance (or for that matter min/max) changes significantly as well.

  • $\begingroup$ Assuming a median could reasonably be calculated, it would flatten the curve and possibly eliminate an otherwise obvious trend. Extreme values are better preserved in the mean temperature, and they are important markers for a trend in the development. $\endgroup$ – user18607 Jan 20 at 12:25
  • $\begingroup$ If I understand correctly, you're saying temperature distribution may NOT be normally distributed, so mean and median may differ. Is that correct? Also, would you still geographically weight temperatures based on station clustering to find the median? $\endgroup$ – Barry Carter Jan 20 at 15:00
  • $\begingroup$ purplemath.com/modules/meanmode.htm $\endgroup$ – trond hansen Jan 20 at 18:06
  • $\begingroup$ The median is robust and resistant, thus useful when data are not normal or when there are outliers. The spread affects median and mean identically as long as the distribution remains normal and the data clean. Do you have any evidence either of those assumptions do not hold for your input data? $\endgroup$ – gerrit Jan 21 at 9:10
  • $\begingroup$ I was just curious about using it since when doing statistics I (not a pro) usually saw differences. Maybe that's an interesting one to see "global weirding" @BarryCarter Temperature is capped from below, so it's perhaps the truncated normal distribution: en.wikipedia.org/wiki/Truncated_normal_distribution $\endgroup$ – Philip Jan 21 at 13:38

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.