let's talk about the non-parametric alternative test to independent samples and the pair samples to test so first we have a menin whitney u test which is the non-parametric alternative to independent samples t-test you run this test in case the nominal assumption is violated and you know the normality assumption for the independent samples and also you run this test when the data in each group is actually indeed ordinal level of measurement remember that for independent samples t-test all the data should be at least interval level of measurement right because you need to be able to calculate the other the mean right but in case your data are in order levels such as like ratings um from like a survey um and then you want to compare their overall rating um say before and i know between like a different groups different independent groups then um you can use man whitney you test to compare them because the man whitney you test is based on the the rank some so indeed it is actually um you know uh designed for the ordinal level of measurement and also it'll actually convert transform the interval or ratio level um of measurement data into the rank in order so that's how it is i'm testing and the two different independent groups so let's just take a look at the um our old data the sample log rpa data to illustrate how to run the meta whitney u test well actually let me just um explain a wilcoxon signed rank w test and then i'm going to show you how to run both bennett whitney and the wilcoxon signed rank test so wilcoxenstein rank test is the non-parametric alternative to pair sample t-test again you use this when the normality assumption is violated then the amount of difference data right and when the data are indeed an ordinal level of measurement so let's just open the data okay so this is the the sample log more va data from with two you can find the data i actually the excel sheet um under the levels of measurement folder so um this is actually um data with you know different um and designs so if you want to compare the visual any visual acuities by gender then it'll become the independent samples t-test so or if you want to just compare the visual acuity between any two um the eyes right the ou versus od versus os or od versus uh multiverses os and you can actually run this as pair samples t test right even though there are different eyes but it is actually um they are belong to the same subject right so we can actually you know consider them as related okay so let's just run the independent samples test first so let's just compare the visual qt the binocular visual qt by gender so click on t-test independent sample t-test now you move all you to the dependent variable and gender to the grouping variable now you want to mean difference descriptive descriptive plots homogeneity test test right so this is the output and if you look at see um you can just quickly see that you know these mean difference will not be statistically significant okay um and as you can see the location of the median is quite far from the mean of each group which suggests that you know these data are not probably normally distributed and so 0. 28 0. 24 so in terms of a mean and it will not be statistically i mean well if there's not much difference um okay you call it the variance and it's okay but we have problem with the normality right all three tests is way less than 0.
05 meaning that the normal assumption is violated right so we cannot use this um okay so p so in this case because the normality is violated we need to take the minute we need it has to see what it is you have to ignore the mean difference and the 95 confidence interval you do not report these values okay because even the developer didn't know how it is possible to actually calculate the mean difference for a minute with the u test it is now calculating the mean difference i mean this function is taken from the r and then it it was not clear you know how this meant with the u test provide mean difference and 95 confidence interval for the mean difference so you should just ignore this only look at this and the p value so this is the manner when you use statistics right and the p value for a minute whitney u test is 0. 339 which is greater than alpha 0. 5 right so what that means is that there is no significant difference um between male and female in terms of their binocular visual okay so this is how you run the manner whitney test and let's run the pair sample t-test so let's just compare binocular versus one monocular the right eye and mean difference descriptive descriptive plots normality test and see [Music] well there's quite a lot of overlap um here so we cannot really see if even though this is not covering the mean of the other but you know if it were to be um normally distributed in a difference then it would not be you know statistical significant but if you look at this the mean va there's almost like a 0.
1 difference more than 0. 1 difference right um so which is kind of a suspicious here we go the test of normality again for parasample t test you only have to check the nominality of difference and all three tests is actually uh telling us uh uh are telling us that the um the normality assumption is violated so we cannot use this right um well even the t test is actually signaling the significant difference but see because the j movie is thinking that these two groups are like independent samples right that's why you have this large error bar so you have to actually adjust accordingly to take account into the design component anyhow because we cannot use this a t-test we need to take wilcox and stand rank test yeah as i expected and the difference is significant i will coxswain's w sandra and fast so this is actual statistics and so it says 23 pairs of values were tied and so basically we have um 23 pairs of having the same log mark visual acuity so if they're rank or tied then you need to do some correction but you don't have to know this um again just to ignore these min difference then there are the difference and 95 confidence interval of the difference all you have to know is this and [Music] i don't know why i believe it comes anyhow so this is the 83 and this is a way less than .