Pretest and Posttest Analysis with ANCOVA and Repeated Measures ANOVA using SPSS

hello this is dr. Grande welcome to my video lysing pretest and posttest results using spss as always if you find this video useful please like it and subscribe to my channel i certainly appreciate it i have here in the data editor in spss fictitious data i'll be using for this example i have an independent variable named treatment and this has two levels 0 and 1 if we go up here to value labels up top we can see this cognitive behavioral therapy CBT and a control group so we have 2 levels of this one independent variable name treatment and then I have dependent variables pretest post-test and also have this post-test three months variable so first I'm just going to be looking at the treatment variable and pretest and posttest so I won't be using this post-test three months variable right now I'll get to that later just pretest and posttest across this between-subjects factor this treatment variable so we have a few choices here and I'm going to be looking at two of them and Cova analysis of covariance and repeated measures ANOVA I'm going to demonstrate both techniques to analyze pretest and post-test data so let's assume in this example that we have this treatment variable we have these two levels so participants are randomly assigned to the CBT group or the control group so this is a between subjects factor and then we have a measure in this case let's say this measure is depression the pretest and the post-test are both both scores both variables are collected using the same instrument an instrument to measure depression levels and the pretest is it ministered before the treatment and the post-test after the treatment so in this case time is a within-subjects factor we have this pretest and then we have a later time post-test so time is within subjects and the type of treatment is between subjects so we're mixing between subjects and within subjects so let's start with an Cova so an Cova looks at whether after adjusting for these pretest observations there is a difference in the post-test means by group so in the case of using in Cova with a pretest post-test design the pretest is the covariant and there are several assumptions that you need to test before conducting an Cova i'm not going to test them here and i will go through them quickly just list them for an Cova you need independence of observations your independent variable in this case treatment needs to have two or more categorical groups in this case it does the dependent variable and the covariant should both be continuous or an SPSS go to the variable view scale an SPSS scale refers to both interval and ratio the interval and ratio level of measurement you want to check the pretest and posttest variables or outliers also these two variables should be normally distributed for each level of the independent variable so be these data for pretest all the CBT data all the control for pretest and the same thing for post-test control and post-test CBT you'll want to check for homogeneity of variance and homogeneity of regression slopes and also the covariant should have a linear relationship to the dependent variable so to conduct and Cova I'm just going to go in to analyze general linear model then univariate this is what the univariate dialog looks like by default the dependent variable in this case will be post-test the fixed factor we the treatment and the pretest will be the covariant that will be down here covariant under plots I'm going to move trip into the horizontal axis and add that let's continue no changes under save they'll go down to options move treatment over to display means for and I'm going to add descriptive statistics estimates of effect size and homogeneity tests continue and click OK so we have the between subjects factors this is the treatment variable CBT and control we can see we have 20 participants in each group we have this descriptive statistics so the mean for CBT control and the total and we see that the mean score for CBT is lower than the mean score for the control group we have the Levine's test we have a non statistically significant result 0.37 two so we would meet the assumption of homogeneity of variance moving down to the tests of between subjects effects we want to interpret the row labeled treatment our independent variable and moving over here to the p-value we can see that it's point one two three so it's greater than point zero five it is not statistically significant so we would fail to reject the null hypothesis and we would assume there is no difference between the groups between CBT and control moving down to the profile plot we can see the score the estimated marginal means on the y axis and we have the two groups here CBT and control so here using an Cova this is one way to look at pretest and post-test data one way to analyze these data and I'm going to go back here to the data editor and look at another method the repeated measures ANOVA so the repeated measures ANOVA takes into account between subjects factor like treatment between subjects factor and they within-subjects factor in this case time the time between the pretest and the post-test unlike an Cova the repeated measures ANOVA could also have more than two within-subjects factors so we could have pretest post-test and this post-test three months variable as well as potentially several other observations we have a post-test at six months at one year so in this first analysis I'm just going to use pretest and post-test and then I'm going to show you what it looks like with all three of these variables pretest post-test and the post-test three months variable for repeated measures ANOVA as was the case for an Cova there are several assumptions that have to be met and again I'm not going to test all those assumptions but it will list them so for PT measures ANOVA you need to have the dependent variables measured at the continuous level so again an SPSS that's scale either interval or ratio for the between-subjects factor in this case the treatment variable you need at least two levels and we have two levels here in this example you want to check for outliers and you want to make sure that your dependent variables are normally distributed for each combination of the levels of the independent variables so in this case we'll just have the one independent variable so it would be the same as it was Frank Cova it would be these data for pretest CBT these data for pretest control the post-test control and the pretest CBT so you want to test that all for these groups the the data and all for these groups are normally distributed you also want to check for homogeneity of variance and see erisa t and we will take a quick look at sarisa t after we run the example that has all three variables all three dependent variables so it's going to started with the repeated measures ANOVA I'll go to analyze again to general linear model except this time instead of univariate and go down to repeated measures and the dialogue that comes up initially is going to ask us to define the number of levels for the within-subject factor and to name that within-subject factor so we know that the name here could be time that is the within-subjects factor and looking at the number levels we're using pretest and post-test so that's two levels and we'll click ad so it's time to and then define and then it brings us to the more traditional dialog that we'd expect and we see we have the within subjects variables available here in this list box one and two so pretest will be one and post-test we to just drag them over to that list box then we have between subjects factors and this looks list box here and that's going to be treatment so we have our between subjects factor treatment CBT and control those two groups as two levels and then our within subjects variables up here under plots I'm going to move treatment to the separate lines text box and time to the horizontal axis I'm going to add that so it's time times treatment let's continue under options I'm going to move treatment time and treatment times time over to the display means for list box I do want to check off compare main effects and for the confidence interval adjustment I'm going to use bonferroni and for display in this display frame at the bottom descriptive statistics estimates of effect size and homogeneity tests let's continue and then okay so again this is the general linear model we have the within-subjects factor the pretest and posttest the between subjects factor CBT and control they are descriptive statistics and we can note here that the pretest scores for that Depression Inventory are very similar fifty three point seven and fifty three point five five and of course they are different for the post-test forty three point six five for CBT and a control group a little higher at forty seven point one moving down to the multivariate tests I'm going to interpret plays trace the top statistic the statistic in the top row for time we can see or the p-value it we have a statistically significant result here for time however for time x treatment the interaction between the within-subjects factor in the between-subjects factor we have 0.155 so it is not statistically significant here note here that we don't have a p-value for mock waste test of sphericity when i run the three dependent variables that will generate a p-value so we'll take a look at that at that time moving down to test Oh within subjects effects again we have the same information the same result here with the rows instead of these statistics plays trays Wilks lambda we have sarisa T assumed and then we have a couple Corrections green house Kaiser and wind felt so in this case it does not matter which one we interpret for time or time times treatment they're all the same so we have statistical significance for time we do not have it for time x treatment and again just like this table here the mock which tests of curiosity this tester within subjects effects is more meaningful when we have three or more dependent variables I'll test that in a moment we have the same results here for the test within subjects contrasts same p-values for the beans for the pretest as 0.181 so we can assume that we have homogeneity of variance there taking a look at Levine's test of equality of error variances for pretest the p-value is point one eight one and for post-test point zero five three so we would fail to reject the null hypothesis in both these cases so we would assume that we have homogeneity of variance for both these variables now moving to the test of between-subjects effects again here we're going to look at treatment and we have a non statistically significant finding here 0.46 now moving down and move down to the profile plot we just have the one and for this value one that's the pretest for - that's the post-test the green line is the control and the blue line is the CBT level of the depend a variable treatment and as you can see I noted before for pretest the CBT and control values are very close as we move over to post-test we see more of a difference the control group had a higher score on that Depression Inventory than the CBT group so for the last analysis I'm going to include this third dependent variable the post-test at three months variable and go back to analyze general linear model and repeated measures we already have this configured from before so I'm going to press reset and again the within-subjects factor name is time the number of levels in this case this includes this last variable so it's pretest post-test and post-test at three months so that's going to be three and then add so time three and define and now we have three within-subjects variables so it's going to be pretest post test and post-test three months and we still have that between-subjects factor the treatment variable I'll move that over to between subjects factor to that list box under plots again treatment two separate lines time up to the horizontal axis click Add and continue then under options I'm going to move treatment and time as well as treatment times time over to the display means for list box compare main effects use a bond throny correction and check off descriptive statistics estimates of effect size and hojae tests let's continue and we're now ready to conduct the repeated measures ANOVA that's okay so we have the three variables now for the within subjects factors pretest post-test and post-test of three months we still have two levels of one independent variable for the between subjects factors and if we take a look at the descriptive statistics we can see for this variable that we added from the last analysis the CBT is a 42 and the control at forty seven point four so the post-test three-month score for control is actually higher than the control score for post-test and the CBT is a bit lower for post-test three months as compared to post-test so moving down to multivariate test again looking at place trace statistical significance there and for time x treatment point zero nine - that's not statistically significant for mock wheeze test of cerissa t we have a statistically significant result here so we violated the assumption of Surrey City so we move down here to the test I within subjects effects we do not want to interpret the Serie City assumed row but rather the greenhouse geyser or the wind felt both of those tests are statistically significant for time but looking at the interaction effect between time and treatment if we were to assume cerissa T and we didn't check for that we would have a statistically significant result point zero four four however of course we did check for stare city and we realize we violated that assumption so if we look at greenhouse geyser or wind felt both of these results are not statistically significant they're both greater than 0.05 then moving down the output we have Levine's test here and we have the same result of course for pretest and posttest as we had before in the prior analysis we just add this third variable and we have a non statistically significant result here 0.34 two so we would fail to reject the null hypothesis we would assume that we've met the assumption of OSA of variance looking at the test of between-subjects effects and this is considering all three of the dependent variables now looking at treatment so this row here the p value is 0.193 that is not statistically significant it's greater than point zero five continue to move down the output I'm going to move to the profile plot again as I mentioned before when looking at the descriptive statistics cbt this blue line the score decreased moving from post-test to post-test of three months and the control group that depression score increased a small amount from the post-test to the post-test at three months I hope you found this video on analyzing pretest and posttest data in SPSS to be helpful thanks for watching

Channel: Dr. Todd Grande

Views: 54,640

Rating: 4.8795562 out of 5

Keywords: SPSS, ANCOVA, RM ANOVA, analysis of covariance, covariance, repeated, measures, repeated measures, normality, sphericity, between-subjects, within-subjects, factor, Levenes, Mauchlys, linearity, outliers, homogeneity of variance, homogeneity of regression slopes, covariate, pretest, posttest, dependent variable, independent variable, outcome variable, ANOVA, significance, variance, outlier, continuous, continuous variable, scale variable, scale, variable, general linear model, counseling, Grande

Id: q_idm_QnVEY

Length: 22min 11sec (1331 seconds)

Published: Thu May 25 2017