Conducting a Repeated Measures ANOVA in SPSS

hello this is dr. brandy welcome to my video and how to conduct a repeated-measures anova when you have a between subjects factor and within subjects factors so in counseling research we use repeated measures ANOVA when we deploy the same measurement the same instrument multiple times with the same participants so looking at these fictitious data I have in the data view here in SPSS you can see I have an ID variable and I have 40 participants in this study and let's assume this study takes place at a community mental health agency and our area focuses anxiety so we have we randomly assign these 40 participants to two levels of one independent variable one is it is an experimental treatment for treating anxiety so this would be the treatment that we developed and want to test out and the other would be the treatment as usual so whatever treatment protocol the community mental health agency would use when trying to help clients that are exhibiting signs of or symptoms of anxiety and then we have pretest so this is given before any treatment occurs then we use the same measurement of anxiety six weeks later and at the end of the treatment which is 12 weeks after it starts we administer the same measurement so this is the same measurement deployed three times for the same participant so these three values represent within subjects the program is between subjects because you have an experimental level and the treatment as usual level so before proceeding with repeated-measures anova it's important to be aware of the assumptions we have to have independent observations the data must be normally distributed and the three in this case the three dependent variables or how however many you have in your study need to be measured at the interval or ratio level what spss refers to as scale so we look at the variable view we can see that those three in this case three dependent variables are all scale and of course the program is nominal because the there's two conditions experimental and treatment as usual and one's not better than the other we can't rank order them they're both simply categories their nominal and of course the ID number is nominal as part of the repeated measures ANOVA procedure we'll also test what's called C ERISA T which is an assumption of repeated measures ANOVA as well but I'll show you how to interpret that result when we get to that point in the analysis so let's start the procedure by going to analyze and then general linear model and then repeated measures and you can see in the repeated measures defined factor dialog we have to declare the number of levels so the within-subjects factor name by default is factor 1 which is fine for our purposes here the number of levels would be the number of dependent variables so that would be 3 and before I add this because this does show up in the output if you want to rename it just so it's a little more clear I certainly can't doesn't affect anything what we're really measuring here instead of factor one and more accurately would be we're measuring the effect of time right of pretest six weeks and 12 weeks so it would be okay to name this time might make it a little easier to read and interpret the results so I'll click Add and you can see now time with the three dependent variables is loaded in the Swiss box and I'll click define down here bottom left and then you can see we need to assign our dependent variables to the within-subjects variables list box and the order is important so we know the pretest needs to be the first and the six weeks would be the second and the twelve weeks will be the third and you can see there's also a list box for between subjects factors in this case we just have one factor and it is program so I'll move that over to the between subjects factor list box looking at the selections here on the right I'm not going to make any changes from model or contrast but in plots for the horizontal axis I'm going to put the program and four separate lines I'm going to put time and I'm going to make sure I add this so it's program times time here and then continue we don't need to make any changes for post-hoc because we only have two levels of the independent variable so there's really nothing that needs to be done here if you had more than two levels then you would want to have post hoc tests for program under save will make no changes and under options we're going to display the means for time and I'll put those over here we want to compare main effects and we do have three dependent variables there so we're going to use adjustment will use a bonferroni adjustment here and it's good to have the descriptives the estimates of effect size and of course the homogeneity tests here I'll click continue and now this dialog is configured for repeated measures ANOVA so I'll click OK and let's take a look at the results so first we had the within subjects factors and you can see there are three we named what was named factor one two time right so we have the pretest one is the pretest two is the six-week test and three is the 12-week post-test those are within subjects factors then we have a between subjects factor of program it just has two levels experimental and trim as usual and you can see the sample size here is twenty for each of them moving down a bit to the descriptive statistics we can see to start out the experimental treatment as usual means on a measure for anxiety so this is the pretest we're fairly close now fifty three point seven and fifty three point five five which is what we'd expect when we randomly assigned participants what we would hope for of course that there's there's gonna be a decrease that occurs as we move to six weeks and in fact you can see in the experimental group it the mean drops to forty three point five and the treatment is usual as well 249 and then for the 12-week measurement we have even more of a drop to 37.4 and the treatment as usual drops to forty five point seven it's worth noting here and we'll see it on the plot that the post-test score the post-test mean that for treatment as usual is forty five which is actually higher than the six-week measure for experimental so that's that's noteworthy then taking a look at boxes test so we need to make sure that we have a good result here for boxes test and you see the significance is point zero zero four but the value are going to compare it to here is point zero zero one so it's greater than point zero zero one we've met that assumption so we can assume that the observed covariance matrices of the dependent variables are equal across groups then moving to multivariate tests I'm going to interpret Wilks lambda you see there are four choices here but we're going to use Wilks lambda so in terms of the effect of time we do have a statistically significant result there and also time times program we have a statistically significant result we're using a an alpha of 0.05 and of course point zero three five is below 0.05 so we have significance for both time and time times program now taking a look at maquas test of sphericity and spirity is where the variances of the differences between all possible pairs of groups are are equal and what's of most interest to us here is the significance value the Alpha is point zero five we want to be greater than that to meet this assumption and point three three 2 is greater than point zero so we have met the assumption of curiosity moving down to test of within-subjects factors again we can assume cerissa T so we'll interpret this line and we have significance for time and time times program we also have a statistically significant result so again to statistically significant results here now we move down the tests within subject contrasts we can see for time and time times program still both are statistically significant then we take a look at Levine's test of equality of error variances so this is a test for homogeneity of variances and for pretest we're above point zero five so we're good and for post-test we are above point zero five but for the six-week test we are below point zero 5 so we have violated the homogeneity of variances assumption here with only one result being significant and us meeting the other assumptions this would be something that I would note I wouldn't let it stop the analysis will still interpret the results but this is something I would note in the limitations so now we move down to test of between-subjects effects so we're looking at what effect did the program have meaning experimental or treatment as usual and the effect was statistically significant at point zero four four close to the alpha level but still below it and the partial a two squared this would be interpreters ten point three percent so ten point three percent of the variance in the dependent variables can be explained by program then moving down here we have estimates and pairwise comparisons so this is measuring time remember so one was the pretest two to six weeks and three to twelve weeks you can see between each pair we have a statistically significant difference and then moving down to the multivariate test course will clamp ax is significant and then we move down to the profile plots so as you look at this plot this is a particularly useful way to understand what's happening with these data for the experimental level of the independent variable and the treatment is usual level we can see that for a time one which would be pretest that they're very close to the same and again that's what we would expect if the participants are randomly assigned would not expect a large difference there but look what happens for the 6-week which is in green here two and the twelve week which is a tan line you can see that with the Green Line here the six week the treatment as usual there was a decrease but there was a much greater decrease here for the experimental as I mentioned before when looking at the means you can see this value at the six week mark for experimental is actually lower then the 12-week for the treatment as usual but again at the 12-week four treatments usually we did see a decrease again so the treatment as usual did work you did see a decrease in the average anxiety scores over time but we saw a greater decrease with the experimental level of the independent variable I hope you found this video on conducting and interpreting repeated measures ANOVA 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

Channel: Dr. Todd Grande

Views: 139,283

Rating: 4.9027948 out of 5

Keywords: SPSS, RM ANOVA, Repeated Measures ANOVA, Levene’s Test, Mauchly’s test, Sphericity, between-subjects factor, within-subjects factors, Homogeneity of Variances, Homogeneity of variance test, normality, Post Hoc, independent variables, dependent variables, Pillai’s Trace, Wilk’s Lambda, Box’s M, covariance matrices, counseling, Grande, Analysis Of Variance, Repeated Measures Design, Statistics (Field Of Study), Mixed ANOVA, Mixed Factor, Mixed Factorial

Id: vhLS1yPax6M

Length: 14min 45sec (885 seconds)

Published: Tue Aug 25 2015