There currently exist several “black box” software libraries for the automatic forecasting of time series. Popular among these are the 'forecast' and 'bsts' packages for R, which have

functions to automatically fit several common classes of time series models, such as the

autoregressive integrated moving average (ARIMA) and the family of exponential smoothing

models, among others. It is often the case that what one gains from the ease in fitting these

automatic methods comes at the cost of predictive performance. In this paper, we propose

several methods to improve the prediction accuracy of automatic time series forecasting, all of

which relate to creating ensembles of models automatically fit from these packages. We

explore different ways that one can construct these ensembles and evaluate each on a

benchmark time series dataset. In addition, we provide the R code used to construct these ensembles.

Orthographic precision refers to the specificity with which letters are assigned to positions within words. In some theoretical models of visual word processing, precision is the optimal end state of a lexical representation; the associations between letters and positions are initially approximate and noisy, but they become more precise as readers gain exposure to the word. In other models, flexible orthographic codes that allow for rapid access to semantics are the optimal end state, and precise representations are only relied upon under specific circumstances. I will introduce a series of event-related potential transposed-letter priming studies designed to test the predictions of these opposing models and determine the functionality of precision within the orthographic system. More specifically, I examine the lexical properties that modulate precision within readers and the skills that modulate precision across readers. In Chapter 2, I demonstrate that words from high-density orthographic neighborhoods that are easy to confuse are encoded more precisely than words from low-density orthographic neighborhoods. In Chapter 3, I demonstrate that precision differs across languages in two groups of bilinguals with different levels of proficiency. In Chapter 4, I compare orthographic precision across hearing and deaf readers to examine how access to phonology contributes to orthographic tuning. However, the results indicate that there were negligible differences between groups. Overall, the results are consistent with a dynamic orthographic system in which precision varies systematically but does not necessarily characterize more developed representations.

The epidermal growth factor receptor is known to be overexpressed in numerous solid tumor types and has been the subject of extensive therapeutic development efforts. Much of the research on EGFR is focused on protein dynamics and downstream signaling, however few studies have explored how the gene is regulated transcriptionally. Here, we identified two novel enhancers (CE1 and CE2) present within the first intron of the EGFR gene in models of glioblastoma (GBM) and head and neck squamous cell carcinoma (HNSCC). CE1 and CE2 contain open chromatin and H3K27Ac histone marks, functionally enhance transcription in reporter assays, and interact with the EGFR promoter. Genetic deletion of CE1 and CE2 by CRISPR/Cas9 editing significantly reduces EGFR transcript levels, with double deletion exercising an additive effect. Similarly, targeted repression of CE1 and CE2 by dCas9-KRAB targeting demonstrates repression of transcription similar to that of genomic deletion. We identify AP-1 transcription factor family members in concert with BET bromodomain proteins as candidate modulators of CE1 and CE2 activity in HNSCC and GBM through de novo motif identification and validate their presence in these enhancers. Genetic inhibition of AP-1 or pharmacologic disruption of BET/AP-1 binding results in downregulated EGFR protein and transcript levels, further confirming a role for these factors in CE1 and CE2. Our results identify and characterize these novel enhancers, shedding light on the role that epigenetic mechanisms play in regulating EGFR transcription in EGFR-dependent cancer types.

The Twenty Statements Test (TST) is widely used in cross-cultural psychology to elicit descriptions of the self-concept through free-format responses. This study examines whether the TST elicits descriptors that are most descriptive of the self-concept. Members of four ethnic groups in the United States participated, to assess the generalizability of the obtained patterns. Participants generated self-descriptions for the actual, ideal, and ought selves, then rated each description for its descriptiveness. Although a large proportion of self-descriptions were rated as “extremely descriptive,” some participants did not use the “extremely descriptive” rating for any of the descriptions they generated. Results suggest that descriptors generated earlier in the sequence are most descriptive, as are those generated in the actual self measure. The ratings of the extent of descriptiveness of the responses did not vary across four ethnic groups in the United States. These results are discussed in terms of the interpretation of TST-generated self-descriptions in cross-cultural research and other potential factors that influence which descriptors are elicited.