SPSS

For every statistical test, use = .05 1. A developmental psychologist conducted an experiment to determine if watching TV violence affects childrens level of aggressiveness and if the type of character (human vs cartoon) performing the violence has any impact. Twenty children watched 2 hours of television that had a high degree of violent content, whereas another 20 children watched 2 hours of television that had a fairly low level of violent content. For half of the children in each of these groups, the main actor in the programming was a human, whereas for the other half, the main actor was a cartoon. Afterward, experts observed the children in a variety of circumstances and rated their level of aggressiveness on a scale where 1 = very nonaggressive and 9 = extremely aggressive. A 2 (violence level: high versus low) X 2 (character type: human versus cartoon) ANOVA was conducted. Below is the SPSS output from this analysis. Use it to answer the questions that follow.
Tests of Between-Subjects Effects Dependent Variable: aggression Source Corrected Model Intercept VIOLENCE CHARACTE VIOLENCE * CHARACTE Error Total Corrected Total Type III Sum of Squares 21.875a 319.225 7.225 9.025 5.625 15.900 357.000 37.775 df 3 1 1 1 1 36 40 39 Mean Square 7.292 319.225 7.225 9.025 5.625 .442 F 16.509 722.774 16.358 20.434 12.736 Sig. .000 .000 .000 .000 .001

a. R Squared = .579 (Adjusted R Squared = .544)

1. VIOLENCE Dependent Variable: aggression VIOLENCE high low Mean 3.250 2.400 Std. Error .149 .149 95% Confidence Interval Lower Bound Upper Bound 2.949 3.551 2.099 2.701
2. character Dependent Variable: aggression character cartoon human Mean 2.350 3.300 Std. Error .149 .149 95% Confidence Interval Lower Bound Upper Bound 2.049 2.651 2.999 3.601

3. VIOLENCE * character Dependent Variable: aggression VIOLENCE high low character cartoon human cartoon human Mean 2.400 4.100 2.300 2.500 Std. Error .210 .210 .210 .210 95% Confidence Interval Lower Bound Upper Bound 1.974 2.826 3.674 4.526 1.874 2.726 2.926 2.074

Estimated Marginal Means of aggression

4.5

4.0

Estimated Marginal Means

3.5

3.0

VIOLENCE
2.5 high 2.0 cartoon human low

character

Simple-effects tests: comparing low- and high-violence programming for each character type:
Univariate Tests Dependent Variable: aggression character cartoon human Sum of Squares 5.000E-02 15.900 12.800 15.900 df 1 36 1 36 Mean Square 5.000E-02 .442 12.800 .442 F .113 28.981 Sig. .738 .000

Contrast Error Contrast Error

Each F tests the simple effects of VIOLENCE within each level combination of the other effects shown. These tests are based on the linearly independent pairwise comparisons among the estimated marginal means.

a. Using correct APA format, write out the results of the test of the violence-level independent variable. What specifically do the results of this test tell you about the effect of watching high- versus low-levels of TV violence on childrens aggressiveness? There is a main effect of the violence-level independent variable. The violence-level of the television programming affected how aggressively children behaved afterward, F(1, 36) = 16.358, p .05. Specifically, children who watched television programming with a high degree of violence (M = 3.25) behaved more aggressively afterward than those children who watched television programming with a low degree of violence (M = 2.40). OR These results tell me that if we ignore character type (i.e., average over it), then children who watched highly-violent TV behaved more aggressively than children who watched TV with only mild violence. b. Using correct APA format, write out the results of the test of the character-type independent variable. What specifically do the results of this test tell you about the effect of watching human characters versus cartoon characters on childrens aggressiveness? There is a main effect of the character-type independent variable.

The character type used in the television programming (human versus cartoon) affected how aggressively children behaved afterward, F(1, 36) = 20.434, p .05. Specifically, children who watched television programming with a human character (M = 3.30) behaved more aggressively afterward than those children who watched television programming with a cartoon character (M = 2.35). OR These results tell me that if we ignore violence level (i.e., average over it), then children who watched human characters on TV behaved more aggressively than children who watched cartoon characters on TV. c. Using correct APA format, write out the results of the test of the interaction between the violence-level and character-type independent variables. Based on the results of this test and the graph of means, what conclusions would you draw about the effects of violence level and character type on childrens aggressiveness? There is an interaction between the violence level and character type independent variables. There was a statistically significant interaction between violence level and character type on childrens aggressiveness, F(1, 36) = 12.736, p .05. Because we have a statistically significant interaction, it means the effect of one independent variable on childrens aggressiveness depends on the level of the other independent variable. In this specific experiment, it means that the effect of television violence (high vs low) on childrens aggressiveness depends on which type of character (human vs cartoon) is performing the violence. From looking at the graph of means, it appears that when cartoon characters engage in highlyviolent acts or moderately-violent acts, children do not behave very aggressively. In other words, no matter how violently the cartoon character behaves, children appear to react in a fairly nonaggressive fashion. However, when human characters engage in highly-violent acts, children react more aggressively compared to when human characters engage in only moderately-violent acts. This is what the pattern of the means suggests. We need multiple-comparison procedures to confirm which individual cell means are different from one another. d. Why would we want to conduct simple-effects tests for this problem? We obtained a statistically significant interaction. When that happens, it is advisable to conduct simple-effects tests to understand why that interaction is occurring. e. Explain/interpret the results of the simple-effects tests. Report your interpretation in correct APA format. When the actor is a cartoon character, children react equally (non)aggressively in response to highly-violent acts (M =2.4) and moderately-violent acts (M = 2.3, F(1,36) = .113, p > .05). In other words, there is no violence-level effect when the actor is a cartoon character. When the actor is a human character, however, children react more aggressively in response to highly-violent acts (M =4.1) compared to only mildly-violent acts (M = 2.5, F(1,36) = 28.981, p .05). In other words, there is a violence-level effect when the actor is a human character. 2. A perception researcher was interested in determining the effects of temperature and relative humidity on perceptions of warmth. Seven participants were randomly assigned to each condition of a 2 (Humidity: 30% vs 70%) X 3 (Temperature: 70 degrees vs 80 degrees vs 90 degrees) experiment.

Participants sat in one of these 6 weather environments for 10 minutes and then rated how hot they perceived it to be on a scale where 1 = Mildly warm and 9 = Extremely hot. The relevant SPSS output is shown below.
Tests of Between-Subjects Effects Dependent Variable: perceived warmth Source Corrected Model Intercept TEMP HUMID TEMP * HUMID Error Total Corrected Total Type III Sum of Squares 103.905a 1761.524 31.476 69.429 3.000 24.571 1890.000 128.476 df 5 1 2 1 2 36 42 41 Mean Square 20.781 1761.524 15.738 69.429 1.500 .683 F 30.447 2580.837 23.058 101.721 2.198 Sig. .000 .000 .000 .000 .126

a. R Squared = .809 (Adjus ted R Squared = .782)

1. tem perature Dependent Variable: perceived warmth temperature 70 80 90 Mean 5.571 6.214 7.643 St d. E rror .221 .221 .221 95% Confidenc e Interval Lower Bound Upper Bound 5.124 6.019 5.766 6.662 7.195 8.091

2. humidity Dependent Variable: perceived warmth humidity 30 70 Mean 5.190 7.762 Std. Error .180 .180 95% Confidence Interval Lower Bound Upper Bound 4.825 5.556 7.396 8.128

3. temperature * humidity Dependent Variable: perceived warmth

Multiple Compari sons 95% Confidence Interval Post Hoc Tests : temperature

temperature humidity Mean Lower Bound Upper Bound Dependent Variable: perceived warmth Std. Error 70LS D 30 4.000 .312 3.367 4.633 70 7.143 .312 6.510 7.776 80 30 4.224 5.490 4.857 Mean .312 95% Confidenc e Int erval 70 7.571Difference .312 6.938 8.205 (I-J) (I) temperature (J) temperature St d. E rror Sig. Lower Bound Upper Bound 90 30 6.714 .312 6.081 7.348 70 80 -.64* .31 .047 -1. 28 -9. 57E -03 70 90 8.571 .312 7.938 9.205
80 90 70 90 70 80 -2. 07* .64* -1. 43* 2.07* 1.43* .31 .31 .31 .31 .31 .000 .047 .000 .000 .000 -2. 70 9.57E-03 -2. 06 1.44 .80 -1. 44 1.28 -.80 2.70 2.06

Post-hoc tests

Based on observed means. *. The mean differenc e is significant at the .05 level.

Estimated Marginal Means of perceived warmth

9

Estimated Marginal Means

humidity
4 30 70 80 90

3 70

temperature

a. Using correct APA format, write out the results of the test of the humidity-level independent variable. What specifically do the results of this test tell you about the effect of humidity on perceptions of warmth? There is a main effect of the humidity independent variable. The level of relative humidity does have an effect on perceptions of warmth, F(1, 36) = 101.721, p .05. More specifically, those who experienced a 70% relative humidity environment perceived more warmth (M = 7.762) than those in a 30% relative humidity environment (M = 5.190). These results tell me that ignoring temperature (i.e., averaging over it), those in a 70% humidity environment believe it is warmer than those in a 30% humidity environment. b. Using correct APA format, write out the results of the test of the temperature independent variable What specifically do the results of this test tell you about the effect of temperature on perceptions of warmth? There is a main effect of the temperature independent variable. The temperature level does have an affect on perceptions of warmth, F(2, 36) = 23.058, p .05. Specifically, these results tell me that ignoring humidity level (i.e., averaging over it), there is a difference in how warm people perceive it to be depending on the temperature level. That is, at least two of the marginal means of the temperature variable differ from each other (M = 5.571 for 70 degrees, M = 6.214 for 80 degrees, M = 7.643 for 90 degrees). c. Using correct APA format, write out the results of the test of the interaction between the humidity-level and temperature independent variables. Based on the results of this test and the graph of means, what conclusions would you draw about the effects of temperature and humidity on perceptions of warmth? There is no interaction between humidity and temperature on perceptions of warmth, F(2, 36) = 2.198, p > .05. Because we DID NOT OBTAIN a statistically significant interaction, it means the effect of one independent variable on perceptions of warmth DOES NOT depend on the level of the other independent variable. In this specific experiment, it means that the effect of humidity (30% vs 70%) on perceptions of warmth is basically the same when the temperature is 70 degrees, 80 degrees, or 90 degrees. From looking at the graph of means, it appears that when there is 70% relative humidity, individuals perceive it to be substantially warmer than when the relative humidity is 30%. This seems to be about equally true at each level of temperature (i.e., it is true at 70, 80, and 90 degrees). d. The only multiple-comparison procedure I performed was a post-hoc (LSD) analysis on the temperature independent variable. Interpret the results of this post-hoc procedure using correct APA format. Why did I not conduct a post-hoc analysis on the humidity independent variable? The protected t (LSD) post-hoc tests show that when the temperature is 90 degrees, participants perceive it to be warmer (M = 7.643) than when it is 80 degrees (M = 6.214; p .05, two-tailed) or 70 degrees (M = 5.571; p .05, two-tailed). Also, when the temperature is 80 degrees, participants perceive it to be warmer than when it is 70 degrees (p .05, two-tailed).

We conduct post-hoc tests to pinpoint differences among marginal means when we have a statistically significant main effect. The main effect of humidity was significant. However, it only has two levels 30% and 70%. Because a significant main effect tells us that there is a difference somewhere among the levels of an independent variable, we can conclude immediately that the two levels are different from each other (because there is only one difference that could be significant).