In order to get the final D-Score, do I need to get out the outliers from the raw data manually or does the program do it automatically
It depends a bit on what method you are using. For the IAT for example, SoSci Survey follows the outlier correction as specified in the respective papers (see the manual for which paper applies to wich variant).
and provides me with the final valid D-score?
Yes, SoSci Survey will store a D-score in the data set.
"Before evaluating, check for extreme outliers that may need to be removed from the analysis"
Such outliers can also occure in the D-score.
In the original cvs output it says in column U "Testergebnis nach Greenwald et al. 2003 (conventional)" and in column U "Testergebnis nach Greenwald et al. 2003 (improved, D-Score)". I believe that I should take the "improved" outcome, right?
Please read the respective paper to understand the differences. By and large, Greenwald et al. (2003) recommend their new "improved" algorithm.
However, I am wondering why do you refer in both types to the same article.
If my memory is not wrong, the paper is about a comparison of the algorithms (and a third one).
How do you calculate the overall D-value? Is it correct that I have to sum up the D-values of all individual participants as of column U (Greenwald et al. 2003, improved) and then divide it by their total number? Or in other words: By taking the mean of column U?
Usually, attides are measured on the individual level. The overall mean often has biases (such as the actual choice of stimuli) that result in 0 not being a neutral point. Therefore, working with an overall mean is technically possible, but may have methodological issues.
In Greenwald et al. 2003, we cannot find the final formula for the "improved D-value" calculation, where blocks 4 and 7 are combined with blocks 3 and 6.
On page 202 there is:
The conventional algorithm excludes
trials from Blocks 3 and 6, treating them as practice. To
assess the usefulness of these data, separate IAT measures were
computed from Blocks 3 and 6 (practice) and from Blocks 4 and 7
(test).
The paper is a bit cumbersome, but when searching for a while one finds the information :)
Is the standard deviation of the blocks indicated anywhere in the original cvs output? If yes: where?
No, this is not saved as separate variable yet.
Is it correct that any positive D-Value must be interpreted as in favour of the supposed stereotype?
In the IAT you have two dimensions, one of them usually negative/positive. A positive value menas a positive correlation between these dimensions.
So if I first refer on the left to "Male + Success" (block 3 and 4), and later to "Female + Success" (Block 6 and 7) and the outcome is mean D-value of 0.70 it means there is a strong association of male and success, correct?
With reference to the categories as defined in SoSci Survey:
The resulting value is positive if the association between categories A and C or B and D is stronger (less reaction times in block 3/4) than the association A-D or B-C (higher reaction times in block 6/7).
Could you explain why the D-value score can be between -2 und +2 but the interpretation by convention only indicates values "below -0,65" and "above +0,65"?
The D-score can even be smaller than -2 or larger than +2, because of the way it is calculated.