Parametric vs. Non-Parametric tests Flashcards
What are the degrees of freedom for one sample t-tests?
N-1
What are the degrees of freedom for paired t-tests?
N-1
What are the degrees of freedom for independent groups t-tests?
(Na - 1) + (Nb - 1)
or
(Na + Nb) - 2
What are the degrees of freedom for pearson’s r test for correlation?
N - 2
What are the degrees of freedom for z tests?
0
What N do one sample t-tests depend on?
N-1
What N do paired t-tests depend on?
N-1
What N do independent groups t-tests depend on?
(Na - 1) + (Nb - 1)
or
(Na + Nb) - 2
What N do pearson’s r tests depend on?
N*
What N do z tests depend on?
Does not depend on N
For all tests but the z-test (i.e. one sample t-test, paired t-test, independent groups t-test and pearson’s r) we need to look at a different row (distribution) depending on …..?
The sample size
What is the degree of freedom?
It is related to the sample size and tells you which distribution you need to use
It also relates to how much data/information you have, and therefore how good your sample statistics are likely to be (because the bigger the sample size, the better estimate a sample mean is of a population mean)
These tests make certain important assumptions about populations from which data are sampled
What are they?
Parametric tests
What are parametric tests?
Tests that make certain important assumptions about populations from which data are sampled
What are non-parametric tests?
Tests that make far fewer assumptions about populations from which data are sampled
These tests make far fewer assumptions about populations from which data are sampled
What are they?
Non-parametric tests
Which test (parametric or non-parametric) can be applied more readily and is the “play safe” option?
Non-parametric tests
What are the parametric testing common assumptions? List 3
1) Populations from which samples are drawn should be normally distributed
2) Variances (standard deviations) of the populations should be approximately equal
3) No extreme scores (since these have a big impact on the estimated sample statistics)
Why use parametric testing if there are so many assumptions?
Because it’s typically more powerful and sensitive than other approaches
Non-parametric approaches are less likely to find more subtle but statistically significant effects in noisy data
Why do we use non-parametric tests?
They impose fewer assumptions on underlying data (you can use it more readily); our data does not need to meet all assumptions of the parametric test and can be not-normal
What do non-parametric tests do?
They throw away information and the emphasis tends to be on ranks of data rather than actual scores
They are less sensitive to potential stat. effects that are present
Why use non-parametric testing if it’s less powerful than parametric?
Because sometimes the assumptions of parametric testing are violated
It is a “play safe” option; it’s for when you don’t have good evidence to support the assumption or you are unsure whether you data are normal or not
What is considered the basis of several non-parametric tests?
Ordering and ranking data
Relative to parametric tests, non-parametric tests rely upon (……….) assumptions about the distributions from which data are drawn
Fewer
Consequently, we would typically use non-parametric tests when it is (……….) to use a parametric alternative because of an assumption violation.
Inappropriate
The disadvantage of non-parametric tests is that they are often (…………) that their parametric counterparts due to throwing away metric information about the (………..) and focusing on (………..)
1) Less sensitive
2) Scores
3) Ranks
You have 2 samples of data, one from population A and the other from population B. The sample from population A has size nA. The sample from population B has size nB.
What are the degrees of freedom (df) for the following tests based on these samples:
1) A 1-sample t-test to assess whether the mean of the sample from population A is different to some hypothetical value muH
1) Na - 1
You have 2 samples of data, one from population A and the other from population B. The sample from population A has size nA. The sample from population B has size nB.
What are the degrees of freedom (df) for the following tests based on these samples:
2) Assuming the samples from populations A&B can be paired and nA = nB = n, a paired t-test is to assess the evidence that the samples came from populations with different means.
2) N - 1
You have 2 samples of data, one from population A and the other from population B. The sample from population A has size nA. The sample from population B has size nB.
What are the degrees of freedom (df) for the following tests based on these samples:
3) An independent groups t-test to assess evidence that the samples came from populations with different means
3) (Na - 1) + (Nb - 1)
or
(Na + Nb) - 2
You have 2 samples of data, one from population A and the other from population B. The sample from population A has size nA. The sample from population B has size nB.
What are the degrees of freedom (df) for the following tests based on these samples:
4) Assuming the samples from populations A & B represent paired variables for a group of nA = nB = n individuals, a Pearson’s r test of correlation between variables.
4) N - 2
