enfascination - User contributions [en]

Institutions and Organizations WR Scott 2001

2009-10-30T01:43:42Z

67.173.45.221:

Publiher is Sage, I've heard of them. I never found the meat of this book, but I was impressed with its historical material, on both institutionalist economics, sociology and poli sci and organizational . Page 2

It says something about my intersts that the parts that grabbed me were all about economics, and even all about Oliver Williamson.

Great, have notes. [http://worldwideweb.unconventionallylonguniformresourcelocator.com/WRScottInstandOrgs.zip Here are the best excerpts], with an obvious personal bias towards economics

Q270

2009-09-27T01:47:55Z

67.173.45.221:

Course Site:
http://cognitrn.psych.indiana.edu/rgoldsto/courses/q270.html

==comments on HW2==
Most people had only a superficial understanding for why the independent samples design yielded insignificant results where paired samples resolved significant results, though you all were pretty good at writing in a way that made it seem like you knew. You will make excellent scientists. I wrote up a little note below giving the deepest reason that I could manage on why a paired samples design/analysis is stronger than independent samples.

==comment on Paired vs Ind Samples T-Test==
A higher level reason that paired samples trump individual samples is that variances are inevitably higher in the independent sample design because less variance is controlled for in the design. This variance makes it difficult for you to be justified in claiming separation of the means of your experimental conditions from each other. And since the goal of the experiment was to show that your different means were actually different, this is bad.

In one sentence, differences between people and everything else create variation in the data, and the paired samples design controls for much of the variation, allowing you to see subtle effects that get hidden in the noise of the high variation unpaired samples.

Here is another way to intuit why it is better to compare the means and variances of 12 pairs of data points than to compare the means and variances of one sample of 12 to that of another sample of 12.

When squaring numbers, you will notice that numbers become bigger much faster as they grow. 2^2 is 4 (twice as large) but 20^2 is 400 (twenty times as large). Since variance is the square of a value's distance from the average value, this property of squaring means that big differences have more influence in determining the variance than small differences.

The two independent samples will inevitably have more points that are further from their two means than each paired sample will, because there are only two means to fit all of your data to, instead of twelve means, each of which controls for many factors (like individual differences). As a result, the variance must be much higher, making it more likely that the natural variance in the estimates will be able to account for the difference in the means that were caused by your experimental condition (which is bad).

If this additional way of thinking about it makes it easier, good. Otherwise, you have permission to forget it.

==Special Note: Location==
We were very fortunate to get hooked up at the Simon Hall lab, with the musical equipment, and I think I can safely say that we will be meeting there for the rest of the semester.

==Part One of Class==
===Lab One===
[[Lab 1 Notes COGS-Q270 Fall09]]
===Assignment One===
Note: The assignment was written with Mac in mind. Where it asks you to go to preferences, go instead of "Options" under "Edit" and find the Viewer tab from there. That said, the little box should be checked by default. Please provide me with clean printouts, not too much line noise if you can help it.

Q270

2009-09-27T00:48:11Z

67.173.45.221: /* Paired vs Ind Samples T-Test */

Course Site:
http://cognitrn.psych.indiana.edu/rgoldsto/courses/q270.html

==Paired vs Ind Samples T-Test==
Here is another way to intuit why it is better to compare the means and variances of 12 pairs of data points than to compare the means and variances of one sample of 12 to that of another sample of 12.

When squaring numbers, you will notice that numbers become bigger much faster as they grow. 2^2 is 4 (twice as large) but 20^2 is 400 (twenty times as big). Since variance is the square of a value's distance from the average value, this property of squaring means that big values have more influence in determining the variance than small values.

The two independent samples will inevitably have more points that are further from their two means than each paired sample will, because there are only two means to compare all results to, instead of twelve means. As a result, the variance must be much higher, making it more likely that the natural variance in the estimates will be able to account for the difference in the means that were caused by your experimental condition (which is bad).

If this additional way of thinking about it makes it easier, good. Otherwise, you have permission to forget it.

==Special Note: Location==
We were very fortunate to get hooked up at the Simon Hall lab, with the musical equipment, and I think I can safely say that we will be meeting there for the rest of the semester.

==Part One of Class==
===Lab One===
[[Lab 1 Notes COGS-Q270 Fall09]]
===Assignment One===
Note: The assignment was written with Mac in mind. Where it asks you to go to preferences, go instead of "Options" under "Edit" and find the Viewer tab from there. That said, the little box should be checked by default. Please provide me with clean printouts, not too much line noise if you can help it.