University of Nebraska–Lincoln



Psychology 450 Laboratory Homework #8LabAMPMName______________________Walk-Through: ANCOVAs with Interactions LabDataset ancovalab_data1_141.savThe purpose of the study was to examine the relationship between Practice & Task Novelty upon Performance. Participants were taught a finger dexterity task involving threading nuts onto a series of stationary bolts as quickly as possible for 30 seconds. They were told they would be performing a similar task and would be paid $1 for each nut they completed. They were told they could practice as many times as they liked before the performance testing. When each participant had practiced as many times as they liked, they were randomly assigned to one of three performance tasks: 1) Novel – the performance task was to push a nut onto a series of stationary posts until it clicked, continue to push while rotating clockwise until it clicked again and continue to push while rotating counter-clockwise until it clicked again, 2) Dissimilar – threading nuts onto a series of stationary bolts, but they were reverse-threaded (had to be threaded counter-clockwise – opposite the traditional “lefty-loosey righty-tighty” logic, and 3) Similar – threading nuts onto a series of stationary bolts (like the training task).Preliminary Analyses#1 Perform the ANOVA using Task as the IV & Perf as the DV.F = ________ df = __, _____ MSe = ________ p = ________Fill in the Performance means for each Task group:NovelDissimilarSimilarMean PerformanceDescribe the results (eyeballing the pairwise comparisons): #2 Perform the ANOVA using Task as the IV & Practice as the DVF = ________ df = __, _____ MSe = ________ p = ________Fill in the Practice means for each Task group:NovelDissimilarSimilarMean PracticeIs practice a confound? #3 Get the correlation between Performance & Practicer = ________ p = ________#4 Something new (and a bit informal)…Split the data file by Task – then get the correlation between Performance & Practice for each groupNovelDissimilarSimilarrpMathematically and statistically speaking, an interaction is about a DV-Covariate regression slope difference across the IV groups. However, often if there are DV-Covariate correlation differences there are also DV-Covariate regression slope differences, but it is much quicker to check correlations than slope (just remember to remove the “split cases” after…)#5 So, is there any indication that an ANCOVA may give a different result than an ANOVA for this analysis?Is practice a strong confounding variable? If so, you will want to “control it”…Is it substantially correlated with the DV? If so, you will want to “control it”…Does it appear to be differently related to the DV for different IV groups? If so, you will want to check for an interaction between the IV & Covariate…ANCOVA Analyses#1 Get the mean and std for practice mean _______________ std _______________#2 Paste below the syntax you will use to compute the mean-centered practice#3 Let’s compare the Task groups at 10, 30 & 50 practices. What value of the mean-centered practices variable corresponds with each of these?10 practices _______________ mean-centered practices30 practices _______________ mean-centered practices50 practices _______________ mean-centered practices#4 Paste below the UNIANOVA code to get the ANCOVA – including the corrected group comparisons at 10, 30 & 50 practices.Syntax:Main effect of task: F = ________ df = __, _____ MSe = ________ p = ________Main effect of practice: F = ________ df = __, _____ MSe = ________ p = ________Interaction: F = ________ df = __, _____ MSe = ________ p = ________What is the slope of the performance-practice regression line for the Similar group? ________ Is the slope significant? _______10 practices pairwise comparisons Similar vs. DissimilarSimilar vs. NovelDissimilar vs. NovelMeans ComparisonEMMEANS p-valuer30 practices pairwise comparisons Similar vs. DissimilarSimilar vs. NovelDissimilar vs. NovelMeans ComparisonEMMEANS p-valuer50 practices pairwise comparisons Similar vs. DissimilarSimilar vs. NovelDissimilar vs. NovelMeans ComparisonEMMEANS p-valuer#5 Use the xls program to get the plot of the interaction model – be sure to get your labels and weights in the right places!Copy the plot here#6 Recode Task so that Novel = 3 Dissimilar = 2 Similar = 1 call it task_SDN Rewrite the UNIANOVA code to use this new “IV” so you can get the slope of the perf-practice regression line for the Novel group.Syntax:What is the slope of the performance-practice regression line for the Novel group? ________ Is the slope significant? _______#7 Recode Task so that Novel = 1 Dissimilar = 3 Similar = 2 call it task_NSD Rewrite the UNIANOVA code to use this new “IV” so you can get the slope of the perf-practice regression line for the Dissimilar group.Syntax:What is the slope of the performance-practice regression line for the Dissimilar group? ________ Is the slope significant? _______#8 Write it up, following the example in the handout. Get all the parts…IntroReport of interaction F-test and description of interaction as different group effects for different values of covariateAlternative description of interaction as perf-practice regression slope differences for different groupsReport of Practice main effect and whether it is descriptive or misleadingReport of Task main effect and whether it is descriptive or misleading.Your TurnUsing any of the data sets from class, go find a combination of a DV, 2- or 3-group IV & covariate that give you an ANCOVA with an interaction.How to proceed? Grab a data setPick a DVPick a 2- or 3-group IV (select cases to discard groups if there are more than 2/3)Pick a quantitative covariateRun ANOVAs with DV and Covariate as dependents – looking for effects & confoundsSplit the file on the IV & get DV-Covariate correlations for each group – looking for effects & confoundsRun an initial ANCOVA, using the raw covariate – if there is an interaction with the raw covariate, there will be for the mean-centered covariate (remember not to use the plotting program with an raw covariate!!) – ta da!!!If you found an ANCOVA w an interaction, follow the steps listed below (paralleling the walk-through) If the ANCOVA doesn’t work, try another!Your model must include a significant IV-Covariate interaction !!!Dataset:_______________DV: __________________Group IV: _____________1st group:____________2nd group:____________3rd group:____________Covariate: _____________ANCOVA analyses #1 Get the mean and std for covariate mean _______________ std _______________#2 Paste below the syntax you will use to compute the mean-centered covariate#3 What raw values of the covariate do you want to perform group comparisons (use 3-4)? What value of the mean-centered covariate corresponds with each of these?1st ___________ _______________ mean-centered covariate2nd ___________ _______________ mean-centered practices3rd ____________ _______________ mean-centered practices4th ____________ _______________ mean-centered practices#4 Paste below the UNIANOVA code to get the ANCOVA – including the corrected group comparisons at chosen valuesSyntax:Main effect of group IV: F = ________ df = __, _____ MSe = ________ p = ________Main effect of covariate: F = ________ df = __, _____ MSe = ________ p = ________Interaction: F = ________ df = __, _____ MSe = ________ p = ________What is the slope of the performance-practice regression line for the highest valued group? ________ Is the slope significant? _______1st covariate value pairwise comparisons Group 1 vs. Group 2Group 1 vs. Group 3Group 2 vs. Group 3Means ComparisonEMMEANS p-valuer2nd covariate value pairwise comparisons Group 1 vs. Group 2Group 1 vs. Group 3Group 2 vs. Group 3Means ComparisonEMMEANS p-valuer3rd covariate value pairwise comparisons Group 1 vs. Group 2Group 1 vs. Group 3Group 2 vs. Group 3Means ComparisonEMMEANS p-valuer4th covariate value pairwise comparisons Group 1 vs. Group 2Group 1 vs. Group 3Group 2 vs. Group 3Means ComparisonEMMEANS p-valuer#5 Use the xls program to get the plot of the interaction model – be sure to get your labels and weights in the right places!Copy the plot here#6 Recode Task so that Group 1 3 Group 2 2 Group 3 1 called ______________ Rewrite the UNIANOVA code to use this new “IV” so you can get the slope of the DV-Covariate regression line for the group that was coded 1 & is now coded 3?Syntax:What is the slope of the performance-practice regression line for the highest valued group? ________ Is the slope significant? _______#7 Recode Task so that Group 1 1 Group 2 3 Group 3 2 called ______________ Rewrite the UNIANOVA code to use this new “IV” so you can get the slope of the DV-Covariate regression line for the group that was coded 2 & is now coded 3?Syntax:What is the slope of the performance-practice regression line for the highest valued group? ________ Is the slope significant? _______#8 Write it up, following the example in the handout. Get all the parts…IntroReport of interaction F-test and description of interaction as different group effects for different values of covariateAlternative description of interaction as DV-covariate regression slope differences for different groupsReport of Covariate main effect and whether it is descriptive or misleadingReport of IV main effect and whether it is descriptive or misleading. ................
................

In order to avoid copyright disputes, this page is only a partial summary.

Google Online Preview   Download