Notes by: Antonia, Richard

Overview:

Today:
    Statistical hypothesis testing
        Sampling distrubutions, confidence intervals
    Students T test
Thursday
    Different data
    chi-square tests
Friday
    ANOVA
    

=============

Example theory to be tested: Does Caffeine improve cognitive performance?

We need a working hypothesis to test this. 

An experimental hypothesis makes a predicaiton about the relationship between two or more events or variables. 

How to test: If cognitve performance is increased, students should be better at a) recalling a text or b) faster to finish their homework after a cup of coffee.

Sl. 6
Independent and dependent variables

For Example b):
    Amount of coffee is independent variable
    Time to finish is the dependent variable. 

So, here, we manipulate the independent variable to see its effect. 

Sl.8
Testing against the null hypothesis

The null hypothees states that any results we find will be just due to chance.

So, the null for b) (notated as H_0) would be that coffee has no effect. 

Alternatively, H_1, is that the coffee has an influence.

This is called a two tailed hypotheses, (neutral with respect to the direction of the effect.)
### Like a nekomata - a two tailed Japanese cat.

Not saying if you will be faster or slower, just that there will be a difference.

Coffee speeds up time, H_2, would be one tailed. 

Goal:

To decide between H_0 and H_1

[...] ##Slides will be online later.

[... Decided to only take notes about extra comments that aren't clear. These will be refferred to by slide number....]

==============

Slide 14
## Design Variant: Have the same group of people who didn't have coffee do the homework, come back another day, drink coffee then and then do the homework again.

## It all comes down to the fact that there are always differences between different people on the same day, or the same people on different days.

Sl.16
## μ stands for mean

Sl. 42
Sampling distribution
Distribution of e.g. mean across infinite number of samples. Can approach normal distribution.
(looks something like this. __..-ˆˆˆ-..__)

In social science, the alpha level is normally .05 or .01.
This is an arbitrary level decided by cultural norms. 

===
This came up in the long meandering discussion. http://en.wikipedia.org/wiki/BLEU

Some of the computers in the pyscholing building have SPSS. 

Statistics block course in February: http://www.coli.uni-saarland.de/kvv/detailpage.php?id=1732