hello this is dr. grande welcome to my video on parametric and nonparametric tests oftentimes in counseling research we need to decide between using a parametric or nonparametric statistical test and in this video I'm going to provide you some guidelines to use when evaluating data as appropriate for a parametric test or a nonparametric test so to start with I have a few common parametric and nonparametric tests here parametric tests include independent samples t-test paired samples t-test which is also known as the dependent samples t-test and one-way ANOVA these three are fairly common parametric tests nonparametric tests
a few examples mann-whitney Wilcoxon signed rank and kruskal-wallis test the choice between using a parametric test a nonparametric test comes down to analyzing the data and determining if it's more appropriate if it meets the guidelines for parametric testing or if it's more appropriate for nonparametric testing and sometimes making this distinction is not easy so let's take a look at the parametric tests generally speaking in counseling research we always look to parametric tests first so if we're analyzing data we're trying to side what type of test to use we're going to evaluate that data analyze that
data in terms of can it be used for parametric testing and if it cannot be used for parametric testing we will use nonparametric tests so the assumptions for parametric test would be examined first and different parametric tests have different assumptions but here are some fairly common assumptions you'll see for parametric tests you need random independent samples interval or ratio level of measurement so no nominal or ordinal level variables only interval or ratio SPSS refers to both interval and ratio together as scale so they have nominal ordinal and scale so in the case of using SPSS
you'll be looking for a scale level measurement the data must be normally distributed many parametric tests do not accept outliers many have the assumption of homogeneity of variance and in general the sample sizes need to be larger for parametric test as compared to what would be the minimum for many nonparametric tests parametric tests are generally preferred because they are more powerful and this is a very specific term I'm referring to statistical power meaning these tests are more likely to detect a difference that truly exists parametric tests are less likely than nonparametric tests to make a
type 2 error a type 2 error is when we fail to reject the null hypothesis when in fact the null hypothesis was false a type 1 error is when we reject the null hypothesis but it was in fact true moving on to nonparametric tests as I mentioned before we usually take a look at the assumptions for parametric tests first and if the assumptions are met we use the corresponding nonparametric test and even though nonparametric tests do not have the same number of assumptions they're not as rigid in terms of the assumptions the data have to
meet there are still assumptions with nonparametric tests the first assumption random independent samples is one of the common assumptions for parametric test as well and then looking at some specific nonparametric tests the mann-whitney test the two samples have to have the same shape wilcoxon signed-rank test requires a symmetric distribution and both the kruskal-wallis and Friedman's ANOVA require the same shape and equal variances nonparametric tests are more conservative they have less statistical power they are less able to detect a difference that is truly there a nonparametric test is more likely than its corresponding parametric test to
produce a type 2 error so moving on to some parametric and nonparametric tests I have here in this table a parametric test on on the left and some nonparametric tests on the right and these are the corresponding nonparametric tests so if you're looking to use independent samples t-test and your data do not meet the requirements do not meet the assumptions in many cases you could use a man whitney test that is the corresponding test the nonparametric equivalent of the independent samples t-test similarly the wilcoxon signed-rank test is the corresponding on parametric test for a paired
samples t-test the kruskal-wallis test is the corresponding nonparametric test for the one-way ANOVA and Friedman's ANOVA is the nonparametric equivalent of a one-way repeated-measures anova and of course these are some of the more popular parametric and nonparametric tests there are other parametric and nonparametric tests and remember it's important to check that the assumptions for tests that you're considering using are met whether that test is parametric or non parametric both types of tests have assumptions I hope you found this video on parametric and nonparametric tests to be helpful as always if you have any questions or
concerns feel free to contact me and I'll be happy to assist you
