Throughout the social sciences, tests have been used for two primary - and different - purposes: a) to estimate where an examinee is located on an ability/trait continuum (e.g., intelligence tests), or b) to classify an examinee as either above or below a particular point on the ability continuum (e.g., criterion-referenced tests). From a psychometric perspective, the scoring procedure of the test should reflect the purpose for which the test is being used. From a practical perspective, the administration procedure should be as efficient as possible. The Myers-Briggs Type Indicator (MBTI; Briggs & Myers, 1976) is a personality inventory designed to classify examinees according to four bipolar dimensions. Although the MBTI is quite popular within corporate America, critics have threatened the validity of the MBTI with two seemingly contradictory faults: the test is too unreliable and too long. The purpose of this study was to examine the degree to which using an item response theory (IRT) / Ph. D.

Latham, Saari, and Fay (1980) asserted that their behavioral observation scales (BOS) produce stably high scale reliabilities across rating occasions. Kane Bernardin (in press) argued that this was due to methodological flaws. They faulted Latham et al. first for scale optimization and secondly, for utilizing the same sample of raters when evaluating the stability of their scale reliabilities. Kane and Bernardin hypothesized that rater-specific illusory halo contributed to producing the stably high reliabilities that Latham et al. reported. This study was a test of that hypothesis. Additionally, the temporal stability of halo, leniency/severity, restriction of range, differential accuracy, and internal consistency, and the intercorrelations among four of these rating behaviors were investigated. Particular interest was directed toward the relationship between differential accuracy and the first three measures, as well as to the relationship between two different definitions of halo. Subjects were 274 undergraduate students. Performance ratings of videotaped managers were collected on two separate occasions. Replication samples were manipulated to be either independent of the original sample with respect to both raters and ratees, raters only, ratees only, or not independent at all. Findings were not consistent with the hypotheses. The stability of scale reliabilities does not appear to be affected by idiosyncratic rater biases such as illusory halo. Furthermore, a strong ratee effect was observed in nearly every assessment of the stability of and correlations among the rating behavior measures. A weak positive correlation between halo and accuracy was found, substantiating previous findings, and the two halo measures were found to correlate strongly with one another. / Master of Science

No / To investigate the cross-cultural validity of international Dutch-English comparisons when using the Dutch Rivermead Mobility Index (RMI), and the intra-test reliability and construct validity of the Dutch RMI. METHODS: Cross-cultural validity was studied in a combined data-set of Dutch and English patients undergoing rehabilitation after stroke, who were assessed with the Dutch version of the RMI and the original English RMI, respectively. Mokken scale analysis was used to investigate unidimensionality, monotone homogeneity model fit, and differential item functioning between the Dutch and the English RMI. Intra-test reliability and construct validity were studied in the Dutch patients by calculating the reliability coefficient and correlating the Dutch RMI and the Dutch Barthel Index. RESULTS: The RMI was completed for Dutch (n = 200) and English (n = 420) patients after stroke. The unidimensionality and monotone homogeneity model fit of the RMI were excellent: combined Dutch-English data-set (coefficient H = 0.91); Dutch data-set (coefficient H = 0.93); English data-set (coefficient H = 0.89). No differential item functioning was found between the Dutch and the English RMI. The intra-test reliability of the Dutch RMI was excellent (coefficient rho = 0.97). In a sub-sample of patients (n = 91), the Dutch RMI correlated strongly with the Dutch Barthel Index (Spearman's correlation coefficient rho = 0.84). CONCLUSION: The Dutch RMI allows valid international Dutch-English comparisons, and has excellent intra-test reliability and construct validity.

The services that define our personal and professional lives are increasingly accessed through digital devices, which store extensive records of our behaviour. An individual's psychological profile can be accurately assessed using offline behaviour, and I investigate if an automated machine learning system can measure the same psychological factors, only from observing the footprints of online behaviour, without observing any offline behaviour or any direct input from the individual. Prior research shows that psychological traits such as personality can be predicted using these digital footprints, although current state-of-the-art accuracy is below psychometric standards of reliability and self-reports consistently outperform machine-ratings in external validity. I introduce a new machine learning system that is capable of doing five-factor personality assessments, as well as other psychological assessments, from online data as accurately as self-report questionnaires in terms of reliability, internal consistency and external and discriminant validity, and demonstrate that passive psychological assessment can be a realistic option in addition to self-report questionnaires for both research and practice. Achieving this goal is not possible using conventional dimensionality reduction and linear regression models. Here I develop a supervised dimensionality reduction method capable of intelligently selecting only useful parts of data for the relevant prediction at hand which also does not lose variance when eliminating redundancies. In the learning stage, instead of linear regression models, I use an ensemble of decision trees which are able to distinguish scenarios where the same observations on digital data can mean different things for different individuals. This work highlights the interesting idea that similar to how a human expert who is able to assess personality from offline behaviour, an expert machine learning system is able to assess personality from online behaviour. It also demonstrates that big-5 personality are predictors of how predictable users are in social media, with neuroticism having the greatest correlation with unpredictability, while openness having the greatest correlation with predictability.

abstract: This research addressed concerns regarding the measurement of cyberbullying and aimed to develop a reliable and valid measure of cyberbullying perpetration and victimization. Despite the growing body of literature on cyberbullying, several measurement concerns were identified and addressed in two pilot studies. These concerns included the most appropriate time frame for behavioral recall, use of the term "cyberbullying" in questionnaire instructions, whether to refer to power in instances of cyberbullying, and best practices for designing self-report measures to reflect how young adults understand and communicate about cyberbullying. Mixed methodology was employed in two pilot studies to address these concerns and to determine how to best design a measure which participants could respond to accurately and honestly. Pilot study one consisted of an experimental examination of time frame for recall and use of the term on the outcomes of honesty, accuracy, and social desirability. Pilot study two involved a qualitative examination of several measurement concerns through focus groups held with young adults. Results suggested that one academic year was the most appropriate time frame for behavioral recall, to avoid use of the term "cyberbullying" in questionnaire instructions, to include references to power, and other suggestions for the improving the method in the main study to bolster participants' attention. These findings informed the development of a final measure in the main study which aimed to be both practical in its ability to capture prevalence and precise in its ability to measure frequency. The main study involved examining the psychometric properties, reliability, and validity of the final measure. Results of the main study indicated that the final measure exhibited qualities of an index and was assessed as such. Further, structural equation modeling techniques and test-retest procedures indicated the measure had good reliability. And, good predictive validity and satisfactory convergent validity was established for the final measure. Results derived from the measure concerning prevalence, frequency, and chronicity are presented within the scope of findings in cyberbullying literature. Implications for practice and future directions for research with the measure developed here are discussed. / Dissertation/Thesis / Ph.D. Communication 2012

abstract: The purpose of this dissertation was to develop a Compassionate Communication Scale (CCS) by conducting a series of studies. The first study used qualitative data to identify and develop initial scale items. A series of follow-up studies used exploratory factor analysis to investigate the underlying structure of the CCS. A three-factor structure emerged, which included: Compassionate conversation, such as listening, letting the distressed person disclose feelings, and making empathetic remarks; compassionate touch, such as holding someone's hand or patting someone's back; and compassionate messaging, such as posting an encouraging message on a social networking site or sending a sympathetic email. The next study tested convergent and divergent validity by determining how the three forms of compassionate communication associate with various traits. Compassionate conversation was positively related to compassion, empathetic concern, perspective taking, emotional intelligence, social expressivity, emotional expressivity and benevolence, and negatively related to verbal aggressiveness and narcissism. Compassionate touch was positively correlated with compassion, empathetic concern, perspective taking, emotional intelligence, social expressivity, emotional expressivity, and benevolence, and uncorrelated with verbal aggressiveness and benevolence. Finally, compassionate messaging was positively correlated with social expressivity, emotional expressivity, and uncorrelated with verbal aggressiveness and narcissism. The next study focused on cross-validation and criterion-related validity. Correlations showing that self-reports of a person's compassionate communication were positively related to a friend or romantic partner's report of that person's compassionate communication provided cross-validation. The test for criterion-related validity examined whether compassionate communication predicts relational satisfaction. Regression analyses revealed that people were more relationally satisfied when they perceived themselves to use compassionate conversation, when they perceived their partner to use compassionate conversation, and when their partner reported using compassionate conversation. This finding did not extend to compassionate touch or compassionate messaging. In fact, in one regression analysis, people reported more relational satisfaction when they perceived that their partners used high levels of compassionate conversation and low levels of compassionate touch. Overall, the analyses suggest that of the three forms of compassionate communication, compassionate conversation is most strongly related to relational satisfaction. Taken together, this series of studies provides initial evidence for the validity of the CCS. / Dissertation/Thesis / Ph.D. Communication Studies 2013

An insufficient financial market means that poor individuals cannot access financial capital, making it difficult for them to generate a stable income. Formal banks see these individuals as unreliable customers because of their financial background and see a risk that these potential customers will not repay their loans, which would put the bank at risk. Banks usually use the borrower’s assets as collateral for their loans. Unfortunately, not many of these poor people have any assets. Microloans can solve these problems by opening up the opportunity for financial capital that enables poor people to make the investments needed to create or develop some form of production and thus increase employment. This research was carried out to analyze if the collected variables can determine the repayment ability of those who got a microloan from the Amhara Credit and Savings institution. Since the borrowers received their loans through two different processes, this study divided the data into two groups. Group 1 received their microloan based on a personality test and the individuals in group 2 received their microloan based on group lending. This division is done in order to be able to eliminate that the lending process itself may have affected the repayment ability. This study is based on random sample data from the Amhara Credit and Savings institution. Regression analyses were performed using the STATA-15 software. The results are not entirely consistent with previous studies because some variables did not get the expected outcome linked similar to previous studies. Some of the variables in this study appear to have an effect on the repayment ability, but not all. Thus, the conclusion is that the results are insufficient and further research needs to be made to reject or confirm the influence of the socio-economic factors and structure of the microloan on the repayment ability for Ethiopian borrowers. / En otillräcklig finansmarknad innebär att fattiga individer inte kan få tillgång till finansiellt kapital vilket gör det svårt för dem att generera en stabil inkomst. Formella banker ser dessa individer som opålitliga kunder på grund av deras ekonomiska bakgrund och ser en risk med att dessa potentiella kunder inte kommer att återbetala sina lån, vilket skulle sätta banken i risk. Banker använder vanligtvis låntagarens tillgångar som säkerhet för sina lån. Tyvärr så har inte många av dessa fattiga människor några tillgångar. Mikrolån kan lösa dessa problem genom att öppna upp möjligheten för finansiellt kapital som gör det möjligt för fattiga människor att göra de investeringar som behövs för att skapa eller utveckla någon form av produktion och därmed öka sysselsättningen. Denna forskning genomfördes för att analysera om våra insamlade variabler kan förklara återbetalningsförmågan hos de som fick ett mikrolån från Amhara Credit and Savings Institution. Eftersom låntagarna fick lån genom två olika processer delade vi upp dem i grupp ett, som fick sitt mikrolån genom ett personlighetstest och grupp två, som fick sitt mikrolån via en grupp med andra individer. Detta för att kunna eliminera att själva processen till hur de har fått lånet kan ha påverkat återbetalningsförmågan. Studien är baserad på slumpmässiga provdata från Amhara Credit and Savings Institution. Regressionsanalyser utfördes med användning av Stata 15-programvaran. Resultaten är inte helt i överensstämmelse med tidigare studier, i och med att vissa variabler inte har det förväntade utfallet kopplat till tidigare studier. Några av variablerna i denna studie visar sig ha en påverkan på återbetalningsförmågan, men inte alla. Således är vår slutsats att resultaten är otillräckliga och behöver ytterligare undersökning för att kunna avvisa eller bekräfta denna uppsats variablers påverkan på återbetalningsförmågan.

Previous research suggests as more scores are interpreted, there is a coinciding increase in the chance significant scores will be obtained. Interpretation of the MMPI-2 can involve the analysis of as many as 98 or more separate scores, suggesting the measure has a strong proclivity for producing a high frequency of seemingly abnormal scores amongst normal healthy adults. In the current study the incidence of elevated MMPI-2 scores was simulated for the normal population using Monte Carlo methodology. Interscale correlations from the MMPI-2 restandardization sample were obtained to determine the percentage of the population with N or more seemingly abnormal scores. Simulations were conducted for all scales combined, and for the Clinical, Harris-Lingoes, Content, Content-Component, and Supplementary scales separately at varying T-score cutoffs. 36.8% of normal adults are expected exhibit at least one elevated score on the Clinical scales at 65T. The normal incidence of at least one seemingly abnormal score was 38.3% on the Content, and 55.1% on the Supplementary scales. When all scale groups are considered together, approximately 50% of the normal population has three or more significant scores, and at least seven seemingly meaningful scores are found for one out of five normal persons. These results imply that consideration of a large number of MMPI-2 scales should be conducted with caution, and that high T-score cut-points may optimally increase confidence in the absence of corroborative test scores and extra test data.

The purpose of this study was to evaluate current practice of using profile analysis of the WISC-III in the diagnosis of processing deficit in earning disabled students. The types of processing deficit prescribed by school psychologists were examined. The validity of the profile analysis approach in processing deficit diagnosis was investigated. Statistical profile analysis, Cattel's r(p(k)) coefficient, and cluster analysis were used to determine whether the mean WISC-III subtest profiles differ among different processing deficit subgroups, and to determine whether individuals diagnosed with the same processing deficit have similar WISC-III profiles. This study was conducted on a sample of 134 non-Hispanic White students identified as learning disabled in a school district in the West Coast area. The results found numerous types of processing deficit, which were not limited to the eight types defined by the school district. About 50% of the individuals were identified with just one processing deficit, while the other half had two or more processing deficits in combination. In addition, the most frequently diagnosed processing deficits were deficits in auditory processing, visual processing and sensory-motor integration. The mean group profiles of the two largest groups with Auditory Processing Deficit (APD) and Visual Processing Deficit (VPD) were compared using the statistical profile analysis technique. The results showed that the mean profiles of the two groups are not parallel. Based on Cattel's r(p(k)) coefficient, both APD and VPD individuals resemble their own mean group profile more than the other group. However, only 52.5% of individuals matched the mean profile of the group in which they were diagnosed. A similar finding was found in cluster analysis. Only 50% of individuals could be correctly classified into their processing deficit groups based on profile similarities. These findings suggested that only half of the individuals diagnosed with the same processing deficit have similar WISC-III subtest profiles. Caution should be exercised in using the diagnostic information and the use of profile analysis technique in making processing deficit diagnosis. Future studies should be conducted to determine the validity of the various processing deficit diagnoses, and how they are related to the WISC-III subtests.

published_or_final_version / Community Medicine / Doctoral / Doctor of Philosophy