Huge Global Data Set: How's The World Getting On?

[ness]

Looks pretty interesting. I'll have to bookmark that site.

[Mitch f]

You're right that it takes a little time. Just like most anything else of value.

I know that, and I appreciate Tim posting this. The older I get the more skeptical I am of data. I've seen too many instances where they were way wrong.

[ness]

I hear you Mitch. I think there are more issues in how data is interpreted than data itself most of the time. You can hardly research any issue without being subjected to somebody's spin.

[Tim Smith]

I know that, and I appreciate Tim posting this. The older I get the more skeptical I am of data. I've seen too many instances where they were way wrong.

What Ness said. It's usually the interpretations that bog down.

For instance, if you plot infant mortality against human population growth in this data set, you'll get a positive relationship (more dead babies = greater population growth). There are a lot of ways to interpret that wrong.

[Mitch f]

What Ness said. It's usually the interpretations that bog down.

.

Agreed, but I'm also used to seeing a huge problem with the data collection itself, long before it even gets to the interpretation part.

[Forsythian]

Agreed, but I'm also used to seeing a huge problem with the data collection itself, long before it even gets to the interpretation part.

Agree wholeheartedly. I'm an analytical chemist by trade, and I'm amazed at how often the validity of the measuring tools gets overlooked.

[Tim Smith]

Agree wholeheartedly. I'm an analytical chemist by trade, and I'm amazed at how often the validity of the measuring tools gets overlooked.

You have to know what the error and biases might be, but analytic chemistry and social/ecological data are going to have two very different kinds of error. Biological systems add layers of complexity to the tidy laws of chemistry. That makes them more variable, more dynamic, more prone to change, more logistically difficult, more sloppy (and less vulnerable to slop), more resistant to generalization, and more likely to be idiosyncratic than anything that emerges directly from the fundamental laws of nature.

In the case of any statistical analysis you might do with a data set like this, you're more or less at the level of hypothesis formation. Without controlled experiments, you're just making observations....

....but given the ethical limits on experiments with humans, it's a good thing people are out there trying to compile these numbers....

....so the liberal elite can turn us all into zombified brain eaters through public health initiatives and Tinky-winky brainwashing.

[bfishn]

After playing with Gapminder for a bit, it's apparent that the big news here is the ability of John Q Public to access and analyze really big data in a versatile, streamlined UI. The engine is the Trendalyzer software, originally developed by the Swedish Gapminder Foundation, then acquired by Google in '07. Google offers it's own public data platform using the Trendalyzer engine, called the Public Data Explorer;

https://www.google.com/publicdata/directory

Using the Gapminder version, you can plot any of the ~200 available indicators in the x axis against nearly any other indicator in the y axis. Then you can show the results in any or all countries and across the entire timespan of available records. Once you've made your selection you can put it all in motion to see relationships that normally lie buried in traditional displays of data.

Want to know the historical comparisons of per capita water consumption in the US and Mexico between '65 and today? It's about a 7 click operation. Want to filter that by males or females? 2 more clicks.

It brings to mind my favorite line from 2001 A Space Odyssey; "My God... it's full of stars".

[Mitch f]

HAL: It can only be attributable to human error

[Terrierman]

So. Is questionable data better or worse than no data? I think that largely depends on how many data sources you have, the more sources the less likely error trends would vary far from the expected scatter and the far less likely bias is to be introduced.