Statistics HELP

[Adonis4000]

Please help me...

I am making a GCSE Statistics project and I need to hand in tomorrow my hypothesis along with some other random details, one of them being calculations.

By calculations, i mean like list all the different calculations that I am going to do, such as mean, median, standard deviation etc.

One tiny problem though... The answers to my hypothesis are basically yes or no (I think this is called binary data, not sure though).

WHAT DO I DO?!?!?! I have no idea what calculations are suitable for this sort of data!

 

More details about my project:

Topic:  Music - Can humans tell the difference at higher quality music?

Test: Allow people to listen to clips of the same song at different bit rates to see if they can tell the difference

Hypothesis:

• People can distinguish music at 128 kbps from 320 kbps
• cannot 128 kbps from 256 kbps
• cannot 256 kbps from 320 kbps
• Males and females will have the same results
• Teenagers will be able to tell the difference more frequently than adults

Sample: 32 people (8 teenager males - 8 teenager females - 8 adult males - 8 adult females)

Diagrams:

• Pie chart (correct or not)
• Bar chart - Males/Females
• Bar chart - Teenagers/Adults

Sorry for the long post

[.spider.]

I'd recommend to follow the MUSHRA approach.

https://en.m.wikipedia.org/wiki/MUSHRA

 

You'll get much more information than just true/false like you can see here

http://listening-test.coresv.net/results.htm

 

IMO the hypothesis is a bit paradox

[SpaceGhostC2C]

I'm afraid the sample is way too small for any meaningful conclusion. But if the point is to design the statistics part and the sample is just an example, that's not a problem.

 

Yes, it is binary data. Because of that, the relevant concepts are those of a binomial distribution. Reporting the basic descriptive statistics is easy by following the definitions of a binomial distribution. The percentage of "yes" answers will be your estimate of the "p" parameter.

 

Testing the hypothesis will be much more difficult. The sample size implies that almost everything will be inconclusive, but abstracting from that, there is an issue of dependence: you have to decide whether to treat the different answers as independent of each other. If you think that's a good assumption, you can test each outcome separately. But that implies there is no correlation between whether a person can distinguish 128 and 256, and whether the same person can distinguish 256 from 320. If you think that the answers are correlated (i.e., people who can't distinguish 128 vs 256 are less likely to distinguish 256 from 320), then you need to adjust your hypothesis testing: you are not testing two independent hypothesis, but jointly testing both of them.