Factorial ANOVA - Review Outlines Exam 3 | PSY 340.00, Study notes of Psychology

Download Factorial ANOVA - Review Outlines Exam 3 | PSY 340.00 and more Study notes Psychology in PDF only on Docsity! Review Outline and Review Activity – Exam 3 – Psy 340 Spring 2008 Reminder! Exam 3 is during our scheduled final exam time… Chapter 10: Factorial ANOVA (ignore calculation sections of this chapter…focus on concepts)  Why is this an efficient research design? Examining 2 or more variables’ effect on DV at once  What does an interaction effect mean? Be able to interpret one given the variables’ meaning.  A 2-way ANOVA as an example of factorial design: o 2 possible main effects (1 for IV 1, 1 for IV2) o 1 possible interaction (between IV1 and IV2) o 2x2 tables may help you notice/interpret main effects (seen by marginal means) and interaction effects (seen by pattern of differences in cell means – that is, in direction and/ or magnitude of row differences)  Be able to look at a 2x2 table and have an idea of whether there might be main effects and/or interactions  Alternately, be able to look at a graph and tell whether there is an interaction (is there a cross- over effect? Is the magnitude of the differences in IV2 the same or different for the different levels of IV1)?  Be able to interpret SPSS output for 2-way ANOVA – is the F test for Main Effect 1 significant? For main effect 2? For the interaction? Chapter 11 – Correlation  What is a correlation? What information does it provide?  What are the interpretations of negative and positive correlations?  Calculate a Pearson’s correlation coefficient (r; will be given the formula)  Issues in interpreting correlations: o significance (what does it mean for a correlation to be significant?) o squaring the correlation (r-squared) gives you the proportion of reduction in error (aka proportion of variance in y accounted for by x) o problems interpreting correlation coefficients due to range restriction (what does range restriction mean? When/how does it create problems?) o Interpreting the size of correlations (know Cohen’s guidelines) SPSS Output for Correlations: Be familiar with the interpretation of SPSS correlation output. Specifically, In SPSS output, to find the correlation between x and y, follow the row for ‘x’ to the column for ‘y’ and you will find 3 lines of information:  First line gives the actual correlation (labeled “Pearson correlation”)  Second line gives p value (significance) of correlation. If p < .05, it is significant  Third line gives sample size used to compute the correlation. Chapter 12 – Regression  Distinction between a predictor and a criterion variable (be able to identify which is which if given an example research question)  Raw score Prediction model – this is the model where you calculate a (regression constant) and b (regression coefficient). o What is the interpretation of a (the regression constant) and b (regression coefficient or slope)?  The model is, o ŷ = a + b (x) If given a regression equation, you should be able to identify which number is a and which is b and interpret each. For example, if ŷ = .87 + 2.56(x), know that .87 is the regression constant where the regression line crosses the y axis (interpretation = y equals .87 when x = 0); and know that 2.56 is b, the slope of the regression line (interpretation = we’d expect an increase of 2.56 points in y for a 1 point increase in x).  Regression line – use the regression equation (prediction model) to find predicted y scores for each person who has an x score. For each person, graph (x, ŷ) – that is, their actual x score against their predicted y score. You’ll see that the slope should match b and the y intercept should be equal to a. o You can randomly choose a few x values to put into the regression equation and find the predicted y score that corresponds to each x. Then plot each of those (x, ŷ) points for your regression line. o You can also use (Mx, My) as a point because the regression line will always go through the two means (of x and y).  Proportionate Reduction in Error (PRE) o Conceptually, the PRE tell us how good our regression equation/line is at predicting y scores. It may be the case that the predictor we chose (x) is very accurate and strongly related to y, so it does a good job of predicting y scores (that is, when a new sample is later measured on y, the predicted y scores came very close to actual y scores). It may also be the case that the predictor did not do a very good job of predicting y scores. o We need PRE as an index of how good our regression line is compared to a mean (baseline) model. That is, if we had no predictor to use in predicting a y value for someone, we could always just predict that they would score at the mean of y. In the mean (baseline) model, we would predict that everyone would score at the mean of y. Obviously, this will almost never be the case, so this probably isn’t a great model. But it does serve as a comparison for our regression model. o PRE is based on two components – Sum of Squares Error (SS error) and Sum of Squares Total (SS total). The SS error component is an index of error made by our regression equation/line when predicting y using ŷ. The SS total component is an index of error made by the mean (baseline) model when predicting y using My. o SS error = Σ (y – ŷ)2 o SS total = Σ ( y – My)2 o PRE = SS total – SS error SS total