3 ms·
"The average time between magnitude 8 and larger Cascadia earthquakes is about 240 years." The outliers on that graph make me extremely skeptical of an average
by akkartik 16y ago
"The average time between magnitude 8 and larger Cascadia earthquakes is about 240 years."
The outliers on that graph make me extremely skeptical of an average[1]. Eyeballing it[2], the most common delta seems to be close to 300 years. Doesn't using the mode make the most sense here, from the perspective of highest-probability outcome?
[1] Should we just outlaw the average? Perhaps a real statistician can advise.
[2] Earthquakes:
-7800 -7200
-7100 -7000
-6750 -6500
-6200 -5800
-5500
-5150 -4800
-4450
-3900 -3800 -3750 -3500 -3300 -3100
-2800 -2500 -2400
-2200 -1900 -1600
-1500 -1200
-1100 -900 -800
-500 -300
-100 200
500 600
700 900
1200 1400
1500
1700
Deltas:
600 100 100 250 250 300 400 300 350 350 350 550 100 50 250 200 200 300 300
100 200 300 300 100 300 100 200 100 300 200 200 300 300 100 100 200 300 200
300 200
Frequency distribution of deltas:
50 1
100 9
200 9
250 3
300 12
350 3
400 1
550 1
600 1
- kenjackson 16y agoI actually think the average makes sense here. The mode will give you '300', but you have 18 that have occurred at 100/200.
- akkartik 16y agoI see, makes sense. There's outliers, but not that many and not that far off.
- seancron 16y agoUsing the 1.5 * IQR (Interquartile Range) rule of thumb, there don't appear to be any significant outliers. Q3 = 400 Q1 = 200 IQR = 400-200 = 200 1.5 * IQR = 300 300 + 400 = 700 200 - 300 = 0 (can't have negative years) All of the data points fall within that 0-700 year range.
- btilly 16y agoWhat really matters is the conditional probability of earthquake given the time since the last earthquake. Based on the numbers that you just gave, about half the time when you make it to 300 years, the next one goes within 50 more years. But the current best guess, as documented in http://www.nature.com/news/2010/100531/full/news.2010.270.html http://www.nature.com/news/2010/100531/full/news.2010.270.ht..., suggests that the actual risk in the next 50 years is about 37%.