Split-Plot ANOVA (Mixed-Design Two-Way Repeated Measures ANOVA) in SPSS

hello this is dr. Grande welcome to my video on conducting a split plot anova in spss as always if you find this video helpful please like it and subscribe to my channel I certainly appreciate it I have here in the statistics data editor an SPSS fictitious data that I'll be using for this example and I'll be covering here a special type of ANOVA referred to as a split plot ANOVA sometimes it's referred to as a mixed factor or mixed design ANOVA and it's a particular type of a two-way repeated-measures anova and if the type of has one independent variable that's between subjects in this case this treatment variable CBT for one group and a control group and a within subjects factor another independent variable that has within-subjects factor in this case a pretest and a post-test so that's why it's referred to as a mix design it's mixing a between subjects independent variable with a within subjects independent variable so with these fictitious data you can see I have 20 participants in the CBT group and 20 in the control group and let's assume that we're using an instrument that measures functioning with a higher score representing a higher level of functioning and we're going to administer this instrument before the treatment with the pretest and after the treatment with the post-test so this is a two-way repeated measures ANOVA and it has one between-subjects factor and one within-subjects factor the between-subjects factor in this case has two levels and of course it can have more than two levels and the within-subjects factor also has two levels a pretest and a post-test and the within-subjects factor can also have more than two levels in this case though it's just two by two so before we get into the analysis I want to check for the assumption of normality and looking at the assumption of normality for this design we want to evaluate four different distributions because we have a two by two the two levels of the independent variable treatment and the two levels of the within-subjects factor so it would be these data for one group these data for another then this distribution and this distribution we want to test all four of those distributions to see if they are normal so I go up to analyze descriptive statistics over to explore and we can test for normality here with just one dialog with this Explorer dialog we can move the pretest and the post-test over to the dependent list list box the treatment variable over to the factor list I'm not going to make any changes here under statistics or under options however under plots I am going to uncheck stem-and-leaf and check off histogram and check off normality plots with tests click continue and click OK and we have the results here from this Explorer dialog we're going to take a look at the skewness values for pretest cbt and kurtosis same thing for pretest and control take a look at those values and post that CBT same thing and control CVT we'll also want to take a look at all the histograms produce no before histograms we'd have here again to buy - so we're evaluating 4 distributions but for right now I'm just going to take a look at the tester normality and I'm going to specifically evaluate the Shapiro Wilk and this is evaluated at an alpha of 0.05 so if we have a p-value greater than point zero five in this case I'm just going to assume that we have a normal distribution that I'm meeting the assumption of normality again you want to evaluate the students and kurtosis and the histograms as well but for right now I'm just going to look at these and I have here for the Shapiro Wilk and p-values all these are greater than point zero five for all for the distributions pretest CBT and control and post at CBT and control so I'm going to sue my met that assumption of normality and now I'm going to move into the tube a repeated measures ANOVA so this will be analyzed general linear model repeated measures and you can see we have a within-subjects factor name by default it's going to be factor one I change this to time who have a administration of this instrument before the treatment and after the treatment and because we have before and after that's just two levels so a number of levels it'll be two and then click Add so we have time with two levels I'm going to go here to the bottom left and click define and you have these within subjects variables here to the right I move pretest over to one and post-test over to two again this is within subjects variables down here we have between subjects factors and that'll be the treatment variable as the variable of two levels CBT and control so here to the right I'm only going to make changes under two of these sections plots and options so under plots I'm going to move treatment to the separate lines text box and time to the horizontal axis and press add and it's time times treatment down here under plots let's continue under options and I check off the descriptive statistics estimates of effect size and homogeneity tests I'm also going to move the treatment times time factor over to display means for so this is a factor interaction treatment times time then press Continue and then press ok and the bottom left to run the analysis so here in the output we have within-subjects factors you can see this is listed as dependent variable we have pretest and posttest and we have between-subjects factors we have treatment with two levels CBT and control and then we have the descriptive statistics and we're looking here at the mean so we have pretest CBT at forty five point nine the post-test CBT at fifty three point zero five so it moved quite a bit there and the pretest control for seven point five five and the post-test control forty seven point four so it dropped slightly moving down here two boxes test our quality of covariance matrices we can see that we have a non statistically significant finding here point zero five seven it should be noted here that for the boxes test we usually evaluate an alpha of 0.01 nine point zero five however in that case this of course is still not statistically significant so we would assume that we have met this assumption we would assume we have equal covariance matrices moving down the output I'm going to skip past multivariate tests and we're going to move here to walkways tests of curiosity now we don't have a p-value here it doesn't return a p-value because test will not run when you have only two levels of the within-subjects factor in this case we do pretest and post-test if we had more than two levels of the within-subjects factor we would have a p-value here for a maquas test of Susa t and to meet the assumption of cerissa t we're looking for a p-value greater than point zero five moving down here to the within the test of within-subjects effects you can see for time it's reporting a main effect that's statistically significant and an interaction effect that's statistically significant time times treatment point zero zero one now with this statistically significant interaction effect we really don't know if we have a statistically significant main effect and I'll show you on the plot in a moment why that's the case but we do have a statistically significant interaction effect we know that times time treatment is statistically significant moving down here to Levine's test again here for Levine's test we have two results two p-values one for pretest and one for post-test both of these p-values are greater than point zero five so we're going to assume that we have met the assumption of homogeneity of variances then moving down to this profile plot here at the end you can see that as we move from the pretest which is time one to the post-test which is time to for the CBT group there is this improvement their scores increased by quite a bit they're on this measure of functioning but for the control group as you move from the pretest to the post-test there's a little change it actually decreases slightly so that's why we can't interpret this main effect of time as necessarily being statistically significant we don't know because it looks like that the change could just be a product of which group the participants are in the CBT group had this change quite a bit of change here from just under 46 to 53 and we had a slight drop here for the control so we can't say for sure that we have a main effect for time with these results with this interaction effect I hope you found this video on performing a split plot anova and SPSS to be useful thanks for watching

Channel: Dr. Todd Grande

Views: 33,860

Rating: 4.9649124 out of 5

Keywords: SPSS, Split-Plot, mixed-design, mixed factor, SPANOVA, Two-Way Repeated Measures, ANOVA, two-way, analysis of variance, repeated measures, rm anova, two-way rm anova, within-subjects, between-subjects, factor, independent variable, dependent variable, interaction, main, effect, interaction effect, main effect, within-subjects factor, between-subjects factor, Mauchly’s, compare means, profile plot, mean differences, variable, statistical significance, data, analysis, counseling, Grande

Id: 5cYY3TAQeQw

Length: 11min 9sec (669 seconds)

Published: Sun Jun 11 2017