Summa Week 14 Flashcards
True or false?: Do we use nonparametric procedures when normality cannot be assumed?
true
True or false?: Do we use nonparametric procedures when data can be transformed to normality
false. we use it typically when data CANNOT be transformed to normality
True or false?: Do we use nonparametric procedures when methods based on non-normal distributions are not available or appropriate
true
True or false?: Do we use nonparametric procedures when there is a sufficient sample size to assess the form of the distribution
false. we use it typically when there is NOT a sufficient sample size to assess the form of the distribution
True or false? The precise definition of nonparametric statistics does not vary among statistics textbooks
false. although not completely, there are differences
True or false? Nonparametric statistics = distribution-free statistics
false. they are different yes mistakenly used interchangeably
What are distribution-free tests?
statistical tests that don’t rely on assumptions about the prob distribution of the sample population
e.g. do not require normally-distributed data
What are nonparametric procedures?
inferential statistics that do not rely on population estimation of which the experiment is sampled from
What are rank stats or rank tests?
ordinal data from lowest to highest and assigning them integer values from 1 to the sample size
How do you resolve tied values in rank stats?
assigning them the average of the ranks if there were no ties
e.g. 17, 19, 19, 25, 28 becomes 1, 2.5, 2.5 (both share 2nd and 3rd place, with (2+3)/2 = 2.5)
What are assumptions of parametric methods?
generally assumes a random sample from a normal distribution
What are assumtions of nonparametric methods?
few assumptions about the pop distribution
What is in comparison in regards to the DV for nonparametric methods?
may be measured on categorical, ordinal, interval or ratio scales
What is in comparison in regards to the DV for parametric methods?
measured on interval or ratio scale, DV
What is comparison in regards to sample sizes for nonparametric methods?
can be small
What is in comparison for samples sizes in parametric methods?
often requires large sample sizes (e.g., n > 30 for the central limit theorem, or pairs of 15, etc.) to appeal to normality
What is the equivalent non-parametric test for an independent samples t-test?
a median test or a rank sum test (compares 2 independent samples)
What is the equivalent non-parametric test for a paired samples t-test?
a Sign tests or a Wilcoxon test (2 matched/related-pairs samples)
What is the equivalent non-parametric test for a one-way ANOVA?
a Kruskal-Wallis test (compares more than 2 independent samples)
What is the equivalent non-parametric test for a two-way ANOVA?
a Friedman test
What is the equivalent non-parametric test for a Pearson’s correlation?
a Spearman’s Rank Correlation
What is the equivalent parametric test for a distribution comparison?
there isn’t one. There is only the non-parametric test of Kolmogorov-Smirnov (damn those Russians)
If you have a tiny sample size, what could you do to estimate the population?
resample by bootstrapping
What is bootstrapping?
a stat method for estimating the sampling distribution of an estimator by sampling with replacement from the original sample
What is most often the purpose of bootstrapping?
deriving robust estimates of standard errors and CIs of a population parameter like a mean
Fill in the blanks: _______ is a stat method for estimating the _____________ of an estimator by sampling with _______ from the original sample
bootstrapping
sampling distribution
replacement
Fill in the blanks: the purpose of ________ is to derive ___________ of ____________ and confidence intervals of a ________ ________ like a mean
bootstrapping
robust estimates
standard errors
population parameter
Is this an advantage of nonparametric procedures? It has higher power
hell naw. this is the reason to do parametric methods!
Is this an advantage of nonparametric procedures? it relies on a few assumptions about the distribution of the sampled population
true. it’s less picky than parametric methods
Is this an advantage of nonparametric procedures? it gets a quick answer with little calculation
true. the Sign test is eeeeeeeeasy
Is this an advantage of nonparametric procedures? it uses a random sample from a larger population when it isn’t available
yep, i.e. bootstrapping
Is this an advantage of nonparametric procedures? is can use a small sample size
hell yeah
Nonparametric methods are most appropriate when the sample sizes are ______
small
When the dataset is large (i.e. n > 100), it makes little sense to use nonparametric stats because
when the samples are large, the sample means will follow the normal distribution even if the respective variable is not normally distributed in the population
Is this a disadvantage of nonparametric procedures? it doesn’t describe parameters and becomes difficult to make qualitative statements about the actual difference between populations
trick question. it’s false because although they don’t describe parameters, they make it difficult to make QUANTITATIVE statements
Is this a disadvantage of nonparametric procedures? it throws away information
true, e.g. the sign test
Is this a disadvantage of nonparametric procedures? it has less power than parametric procedures
yep
Each nonparametric procedure has its peculiar sensitivities and blind spots such as: the Kolmogorov-Smirnov 3-sample test is sensitive to differences in the location of distributions as well as is greatly affected by differences in shapes
false. although the statement is true, the Kolmogorov-Smirnov is a 2-SAMPLE test, not a 3-sample one
True or false: the Wilcoxon matched pairs test assumes that one can rank order the magnitude of differences in matched observations in a meaningful manner.
true
If the rank order of magnitude of differences in matched observations is not possible, what should you use?
a Sign test
True or false: it is always advisable to run at least 2 nonparametric tests
true. should discrepancies in the results occur contigent upon which test is used, one should try to understand why some tests give different results
