Grandiose/Vulnerable Narcissism predicting spiteful punishment after self-threat

NPI items were re-coded such that the narcissistic response was always scored as 1, and the non-narcissistic response was always scored as 0. The Vanity, Exhibitionism, Superiority, and Entitlement subscales were positively skewed, and were Log-transformed. Power analyses showed that the desired sample size given a medium effect size, .80 power, and an α of .05, was 168 participants. Given that the dataset was more than double that, in order to avoid inflated p values, the total dataset was split into a training (n = 221) and a validation (n = 233) sample based on a random selection of 50% of the total cases. This serves as an internal replication for performed analyses, as suggested by Hair and colleagues (2009). Initial preference scores were calculated using an algorithm created by LeBel and Gawronski (2009). Hypothesis 1: in the no threat condition, high and low vulnerable narcissistic participants (MCNS) would not differ in their endorsements of spiteful or strategic punishment. However, it was predicted that participants high in vulnerable narcissism in the self-threat condition would favor more spiteful punishment than threatened low narcissistic participants. This was tested using a hierarchical linear regression entering in MCNS and Condition in step 1, and adding MCNS x Condition in step 2. Hypothesis 2: in the no threat condition, participants high in grandiose narcissism would not differ from low grandiose narcissism participants in their endorsement of spiteful punishments. But in the self-threat condition, participants high in grandiose narcissism would show greater endorsement of spiteful punishment compared to participants low in grandiose narcissism. This was analyzed using total NPI scores (Hypothesis 2a) and NPI subscales separately (Hypothesis 2b). This was tested using hierarchical linear regressions entering in NPI measures and Condition in step 1, and adding NPI x Condition interactions in step 2. Hypothesis 3: vulnerable narcissism would predict more of the variance in spiteful punishments than grandiose narcissism. This was tested using Fisher's r to z transformation to analyze correlation coefficients between models. Hypothesis 4: participants higher in vulnerable narcissism would be less in favor of strategic punishment after self-threat compared to those lower in narcissism and non-threatened participants. This was tested using a hierarchical linear regression entering in MCNS and Condition in step 1, and adding MCNS x Condition in step 2. Hypothesis 5: participants higher in grandiose narcissism would be less in favor of strategic punishment after self-threat compared to those lower in narcissism and non-threatened participants. This was tested using hierarchical linear regressions entering in NPI measures and Condition in step 1, and adding NPI x Condition interactions in step 2. Hypothesis 6 threatened vulnerable narcissists should show increased implicit self-esteem after given the ability to punish.自恋人格量表（Narcissistic Personality Inventory, NPI）的条目经重新编码，将符合自恋倾向的作答统一记为1，非自恋倾向的作答统一记为0。其中虚荣性、显露性、优越感与权利感四个分量表呈正偏态分布，故对其进行对数转换。功效分析（Power analysis）结果显示，当效应量为中等水平、检验效力为0.80、显著性水平α为0.05时，所需的最小样本量为168名被试。由于本次数据集的样本量远超该值的两倍，为避免p值被高估，研究人员按照总案例的50%进行随机抽样，将总数据集划分为训练集（n = 221）与验证集（n = 233）。正如Hair及其同事（2009）所建议的，该划分方式可对已实施的分析进行内部重复验证。 初始偏好得分通过LeBel与Gawronski（2009）开发的算法计算得到。 假设1：在无威胁情境下，高、低脆弱型自恋被试（MCNS）在对恶意惩罚与策略性惩罚的认同度上无显著差异。但研究预测，在自我威胁情境中，高脆弱型自恋被试相较于低脆弱型自恋被试，会更倾向于支持恶意惩罚。该假设通过分层线性回归检验：第一步纳入MCNS与情境变量，第二步加入MCNS与情境的交互项。 假设2：在无威胁情境下，高、低夸大型自恋被试在对恶意惩罚的认同度上无显著差异。但在自我威胁情境中，高夸大型自恋被试相较于低夸大型自恋被试，会表现出更高的恶意惩罚认同度。该假设通过两种方式分析：一是基于总NPI得分（假设2a），二是分别基于NPI各分量表得分（假设2b）。检验采用分层线性回归：第一步纳入NPI测量指标与情境变量，第二步加入NPI与情境的交互项。 假设3：相较于夸大型自恋，脆弱型自恋能够解释恶意惩罚得分中更多的变异量。该假设通过Fisher r到z转换，对不同模型间的相关系数进行检验。 假设4：相较于低自恋水平被试与未受威胁的被试，自我威胁后高脆弱型自恋被试对策略性惩罚的支持度更低。该假设通过分层线性回归检验：第一步纳入MCNS与情境变量，第二步加入MCNS与情境的交互项。 假设5：相较于低自恋水平被试与未受威胁的被试，自我威胁后高夸大型自恋被试对策略性惩罚的支持度更低。该假设通过分层线性回归检验：第一步纳入NPI测量指标与情境变量，第二步加入NPI与情境的交互项。 假设6：在获得惩罚权后，受自我威胁的脆弱型自恋者的内隐自尊水平会有所提升。