Linh Nguyen

I view personality broadly as any psychological individual differences that may vary both between and within individuals. Beyond the Big Five taxonomy, I am interested in how traits interact at an individual level to form meaningful profile patterns. Further, I am fascinated by intraindividual variation across time and situation, both as a factor of lifespan development, meaningful situations, and informant perceptions.

Psychological measurement aims to devise methods to assess unobserved constructs and validate these methods using an array of psychometric approaches. Although often not a central focus, measurement is at the heart of psychological sciences. I am most interested in latent-variable modeling and assessing the degree to which measures are operating in the same way across different populations and contexts (measurement invariance) and in relation to systematic response biases.

Selected Projects

1. thirt - An R Package for simulating responses and estimating parameters for the Thurstonian Item Response Theory Model. [Github] and [Master's Thesis].
In high-stakes settings such as hiring and promotion, personality measures are often used to aid decision making. However, traditional Likert-scale measures are prone to response bias or "faking good". Forced-choice questionnaires alleviate this bias by forcing participants to rank multiple options that are potentially equally as desirable. These measures, however, present an analytic problem because rank scores are ipsative and cannot be compared between-people. The Thurstonian Item Response Theory model is a solution. This R package uses a random-walk Metropolis Hastings Markov chain Monte Carlo algorithm. Users are able to simulate responses and estimate parameters for scoring purposes. We further compare the solution and performance with other available softwares.
2. Model-X Knockoff for variable selection: A simulation study and applications in genome-wide association studies. [Report].
Model-X Knockoff is a variable selection method designed to control the false discovery rate. We conducted simulation studies to examine the robustness of this method and compare its performance with other standard methods under both moderate (n > p) and high-dimensional (n << p) settings. We applied the method to analyze high-dimensional genetic data using single-nucleotide polymorphisms to predict height. The results showed that knockoff is a powerful variable selection tool that works under various conditions where the traditional methods fail. However, there is significance computation cost associated with constructing knockoff variables compared to other methods.
3. A deep dive into the Multigroup Ethnic Identity Measure: Psychometric properties and factor structures across large ethnically diverse samples.
Ethnic identity refers to people's sense of self in regard to their ethnic group. The predominant measure for ethnic identity is the Multigroup Ethnic Identity Measure. Although the construct is inherently cultural, there hasn't been a comprehensive review of the measure in an ethnically diverse sample. We conduct a thorough psychometric review of the measure and examine the robustness of these properties using a large diverse sample. This includes: (1) basic psychmetric properties such as scale reliability, mean-level scores, and subscale intercorrelation; (2) convergent and divergent validity with other measures of identity development and well-being; (3) measurement invariance tests across ethnic groups with a focus on effect sizes and confirming the latent factor structure.
4. Differential response patterns to personality measures in first and second languages: Cultural frame switching and cultural accommodation.
Research on the effect of cultural cues, such as language, on response patterns to personality measures of bilingual speakers has produced contradictory results. We included bilingual speakers from four different language pairs and examined whether differences in personality scores between the languages would align with differences in social desirability ratings of those personality traits in their two associated cultures. We found contradictory and weak support for the phenomenon of cultural frame switching or accommodation. Secondly, we looked at aggregated data across all specific languages to examine first-versus-second-language effects. Fluent participants reported more positive personality profiles compared to those with lower reported language proficiency.