Core.ecu.edu
Tips on ANOVA1 AssignmentConducting the AnalysisInterpreting the OutputTo calculate the 2, simply divide the between groups sum of squares by the total sum of squares (yellow fill above). To get the confidence interval, see my document Using SPSS to Obtain a Confidence Interval for R2 From Regression. Here is a screen shot showing how I got the confidence interval for eta-squared. You simply enter the F value, df1, and df2 (pink highlight) into the sav file, run the script, and then look back at the data file. Eta-squared will be in the r2 column and the CI in the lr2 and ur2 columns.NMeanStd. DeviationStd. Error95% Confidence Interval for MeanLower BoundUpper BoundBlack D, Black P2505.991.592.1015.796.19Black D, White P2505.301.524.0965.115.49White D, Black P2505.931.564.0995.736.12White D, White P2505.961.604.1015.766.16Total10005.801.595.0505.705.89ANOVAGuilt Sum of SquaresdfMean SquareFSig.Between Groups82.131327.37711.090.000Within Groups2458.8449962.469Total2540.975999ANOVA Effect SizesaPoint Estimate95% Confidence IntervalLowerUpperGuiltEta-squared.032.013.054Epsilon-squared.029.010.052Omega-squared Fixed-effect.029.010.052Omega-squared Random-effect.010.003.018a. Eta-squared and Epsilon-squared are estimated based on the fixed-effect model.Post Hoc TestsGuiltRyan-Einot-Gabriel-Welsch Rangea GroupNSubset for alpha = 0.0512Black D, White P2505.30White D, Black P2505.93White D, White P2505.96Black D, Black P2505.99Sig.1.000.904Any two means that are in the same subset column are NOT significantly different from each other. For these data the 5.30 is significantly lower than all of the other means. This information is then transferred to an APA-style table like that one below:Table 1Certainty of Guilt Related to Race of the LitigantsGroupMSDBlack Defendant, White Plaintiff5.30A1.40White Defendant, Black Plaintiff5.93B1.56White Defendant, White Plaintiff5.96B1.60Black Defendant, Black Plaintiff5.99B1.52Note. Means sharing a letter in their superscript are not significantly different at the .05 level according to a REGWQ test.The REGWQ procedure was used for pairwise comparisons, capping familywise error at .05. As shown in Table 1, certainty of guilt was significantly less when the defendant was black and the plaintiff white than for any of the other. All other comparisons fell short of statistical significance.Errors Made in the PastTwo point errorsNeglecting to identify the independent variableNeglecting to identify the dependent variable.Declaring the ANOVA to have produced nonsignificant results when the value of p was less than .05.Reporting an inappropriate value of p.Reporting an incorrect value for 2.Reporting an incorrect CI for 2.Not including superscripts in the table of means to indicate which means were significantly different from which others.Using, in the table of means, superscripts that do not correctly indicate the obtained pattern of results from the REGWQ.Not correctly describing in words the pattern of results from the REGWQ.One point errorsNeglecting to indent the first line in each paragraph.Reporting an incorrect value for the error degrees of freedomGetting the direction of effect backwards – for example, saying that a mean of 5.29 was significantly greater than a mean of 5.86.Indicating that ethnic identity was affected by the guilt ratings. This is a reversal of cause and effect. It is the guilt ratings that are affected by the (manipulated) ethnic identities of the litigants.Reporting a trend analysis. A trend analysis is appropriate only when the independent variable is quantitative. The independent variable here (ethnic identities of the litigants) is qualitative.Not including in the table of means a note that explains that a REGWQ was conducted and what the superscripts mean.Not arranging the groups in ascending or descending order of their means.Indicating that Bonferroni tests were conducted – REGWQ tests were conducted.When describing, in words, the results of the REGWQ, not emphasizing the direction of the difference between means.Indicating that a 95% CI was constructed. It should have been 90%. ................
................
In order to avoid copyright disputes, this page is only a partial summary.
To fulfill the demand for quickly locating and searching documents.
It is intelligent file search solution for home and business.