The fact that Atkinson showed up here

[Sonicularity]

19 minutes ago, crenca said:

 

 

This math/probability/outcome is correct.  

 

Also, removing tests of 6 or 7 similar presentations impacts the research, regardless if this data edit is known by the test subjects.   Why allow the test subjects to take a trial where at least 2 of the presentations must be different?  Just make it random and tell everyone it is random.  

[Sonicularity]

11 minutes ago, crenca said:

 

You sure about this?  I thought removing outliers was valid in such things.  I happen to have two close relatives (what are the odds?  ) who are PhD statisticians I could ask if I get motivated to drop one of them an email...not that I am...

 

With statistics, I'm rarely sure.  I'd run the numbers on how often I'm sure, but I couldn't be sure about the results. 

 

In the scenario mentioned above, it seems to me that the assumptions could be altered, but I don't know about the results, but it seems likely.  If I tried a similar test using only 2 presentations, and both samples had to be used for every session, everyone would either get both right or miss both.  The results would be different if the same presentation could be used for any session, where only a single miss could occur.  I think it would change both the assumptions and the results.