Saturday, March 10, 2012

Combining the results of several studies related to - if done right - can improve the statistical weight of its conclusions

When it comes to study the roles of gender and sexuality size really matters, but not in the way you might think. A recent example occurred in the research on gender differences, between 1993 and 2007, involving more than one million people.

general statistical data, the better. It is true that what you do with it makes a difference, but great is the best in the land data. Nothing will put a smile on your face to the statistics, but a strong presence.

But as I explained in a previous article, take a large sample is often simply not possible for whatever reason usually to do with time and money. For example, data collection of over one million people, as in the study of gender differences would be a great feat.

One way to solve this problem is related to a meta-analysis.

The term meta-analysis is so common in newspaper articles as they are often not given an explanation of what it really is. When we read: "Scientists have conducted a meta-analysis", it is assumed that this is a good thing and move on. But there is always a good thing, whatever it is?

prefixing things with the target word certainly makes them look a bit futuristic and cool. "Meta" often means that the thing is to talk to him. Very intellectual. For example, the language is a language for talking about language. Metadata is data about data. You create the metadata every time you mark a photo of a friend on Facebook looking worse for wear after a night.

The study of gender differences has a large sample size to do just that. The scientists combined 834 studies from 87 countries and seven national data sets to give them more than a million topics.

This pooled data set, we hope, will better help us find what is happening in the world and not just in the small population examined in a particular study. But these studies, although related, is different. You can have characteristics slightly different population, the quality and size of samples, researchers may have made different assumptions. Therefore, do not compare apples with apples. More like apples to apples apples.

is tempting to think that the data, just the sharing of studies sufficient to do everything better. Big can often be the best, but "garbage in, garbage out" is true, and this is sometimes overlooked.

0 comments:

Blog Archive