Normality Tests in SPSS

hello this is Dr Grande welcome to my video on normality tests in SPSS so often times in research it's important for us to know or we want to know if distribution is normal and there's couple different ways we can test for that in SPSS that I'm going to show you so here's kind of typical setup in data view in SPSS we have 90 participants each one has an ID we have two groups in this case control and CBT group and then we have some pre-test scores and some post- test scores so let's say in this instance we want to determine if our pre test and post test scores are normally distributed now of course in this example these are dependent variables and when we think of normal distributions we tend to think of independent variables but there are wide variety of situations where we might want to test for normality so first we'll go to analyze then descriptive statistics then explore and you can see already have this populated but by default of course this would be empty and we'll move preest and post test both into the dependent list box and statistics will'll leave the same just the descriptives and plots by default would actually look like this but all we really need for this test would be histogram and normality plots with test the box plots can even be set to none then select continue and then click okay and you can see there's lot of output here for both of these variables first we have the case processing summary in this case had no missing values so this is fairly straightforward then looking at these descriptives you can see there's wide range of descriptives available for each variable want to take special note of the value of skis and curtosis in each example see that it's 078 skewness for the pretest but it's 365 for the post test and for curtosis it's ne33 for the pretest and. n37 for the post test and then we move down little further and we have the actual test or normality there are two test or normality the first one is the korov smaroff test which is usually referred to as KS test and the second is the Shapiro Wilk test and you can see each variable is listed firstus the pre-test it gives you the actual statistic and then the significance so what this significance value tells you is for this variable is the data statistically significantly different from normal distribution and in this case you could see it's 02 if we use the alpha as 05 this would not be statistically significantly different from normal distribution so we would say that the null hypothesis is accepted that is we believe that the preest data is normally distributed and similarly for the Sho wil 724 that is not statistically significant so again we would fail to reject the hypothesis and we would assume that the data in the preest variable is normally distributed now for the post test variable our results are different we have different statistic of course and we can see that there is statistically significant KS test result and statistically significant Shapiro will so in this case because these values are lower than 05 we would reject the null hypothesis and therefore we're saying that the post test data is not normally distributed so you might be wondering if these are both test or of normality why is there such difference between the significance values for the ks test and the significance Val values of The shapir Wil these tests analyze the data in different way and it's fairly commonly accepted that the Shapiro Wilk value is more valid as you can see in this case you would reach the same conclusion in terms of the null hypothesis either way for both of these variables but there certainly could be instances where it appears normal under say the ks model and does not appear normal under Superior wil or the opposite could be true and in that case if you get differing results between the two test of normality go with the Shapiro Wilk so as we move down we can see this is the histogram for the preest which the the Shapiro will and the ks test indicated that this data was normally distributed so this is the histogram it does look more or less normally distributed and this is the histogram for the post test which of course the test of normality indicated that this data was not normally distributed and just by looking at fairly clear it is not normally distributed hope this video on testing for normality was useful to you as always if you have any questions or concerns feel free to contact me and I'll be happy to assist you