The tests would be different for the two cases. I get the feeling that you might be using analogies to try and figure out the question to a real-life question....? If so, it's more effective to just post the actual question.
I watched two people blink three times a day for seven days. The hypothesis is that they blink significantly more in the morning compared to the night time. What stat test would you use? Other than mean and error bars, I got nothin.
Or how about this...I watched a two dogs, one young and one old, to add to the sample size, and to possibly accommodate for age differences, go into a room with three boxes, each a different color. My hypothesis is that dogs choose the green box more than the other boxes. Same stat test for the last one?
The tests would be different for the two cases. I get the feeling that you might be using analogies to try and figure out the question to a real-life question....? If so, it's more effective to just post the actual question.
humanbeing (04-29-2016)
I have a project due soon, but I am very confused, so I will. My hypothesis is that classical music makes dogs bark more than other genres of music do. I have two people, a young dog and an old dog, as my sample size. They were both shown one minute of one genre three times a day for seven days. Seven genres were used. I counted barks. I now have graphs. Other than posting the mean and adding error bars, I am not sure what statistical test to use to find a significant difference in the number of barks i.e. p-values, degrees of freedom.
With only two dogs you can't really make useful statistical inferences about the effects of music on dogs in general. You would need to sample more dogs. I would suggest treating you study as a single case design and simply descriptively graphing your results.
Okay, thank you. Could you tell me what test I would use if I did have a large enough sample size, say 500 young male dogs and 500 old male dogs?
Also, what test would I use to find possible age differences? an unpaired t-test?
If you just wanted to compare frequency of barks by genre, you could use a chi-square test. Unpaired t-test is for continuous data.
If you wanted to simultaneously look at the effects of age and genre on bark frequency, you'd need a method for count data, maybe Poisson regression or something similar.
If I combined the dogs into one column, I don't see how I could use a chi-square test. It would be 1 x 7 column. Could you explain? I did use a chi-square to test for differences between the young and old dogs, so thank you.
A chi-square test can be used to compare observed frequencies vs "expected" frequencies under a null hypothesis. Under the null hypothesis that genre has no effect on frequency of barks, the expected frequencies in your column vector are all the same: Each of the 7 cells would have an expected frequency of (total barks/7). You can then enter the observed frequencies and run the test.
Online app here to run the test if you need it: http://graphpad.com/quickcalcs/chisquared1.cfm
Tweet |