1. Jan 22, 2014

### musicgold

Hi,

This is not really a homework question. Attached is a slide from a statistics presentation I found on the web.

I am not sure what conclusion I can draw from this data if I ignore the outlier (1906). Saying "higher magnitude earthquakes result in fewer deaths" seems totally counteractive.

Thanks.

#### Attached Files:

• ###### corre question - Copy.pdf
File size:
34.4 KB
Views:
77
2. Jan 22, 2014

### Staff: Mentor

It does look funny, but the data are not well organized. The high magnitude low death datapoint is for an area with very sparse population, so should be excluded. A better graph would compare quakes in similar population areas with similar building codes.

3. Jan 22, 2014

### musicgold

Ok. Thanks.

4. Jan 22, 2014

### SteamKing

Staff Emeritus
I think the word phrase you are looking for is 'counter intuitive' rather than 'counteractive'.

5. Jan 22, 2014

### haruspex

Might be able to do a bit better than that. If you knew the density of population in each area you could normalise the data by taking the deaths as fraction of population. Looks like population density is influencing the numbers rather more than the severity of the earthquake is. Question is, what area to take around each site? Ideally, it would be some kind of integral wrt radius from epicentre, something like $\int_{r=0}\frac{density(r)}{1+k r^n}rdr$, but maybe you could just fix on a sufficiently large circle to encompass all deaths.

6. Jan 22, 2014

### AlephZero

Also, you could investigate what size of correlation coefficient is significant, for such a small sample size.

Or even whether the notion of "correlation" is meaningful at all, with so few data points. To take a ridiculously extreme example, if you only have two data points, you will always get a correlation of +1 or -1.

Maybe .... but the relevant building codes would be different in a low population density rural area, compared with skyscrapers in a city center. And earthquakes don't necessarily happen where planners think they are most likely to happen.

The danger of going down this route is that you do a lot of research and end up with 7 different "stories" about 7 different events, but you still can't really draw any general conclusions because there events don't have much in common except they were all "earthquakes".