Dissertation/Thesis Abstract

Scaling Undergraduate Scientific Writing via Prominent Feature Analysis
by Gallo, Katarzyna Zaruska, Ph.D., Mississippi State University, 2018, 150; 10974642
Abstract (Summary)

Prominent Feature Analysis (PFA) is a reliable and valid writing assessment tool, derived from the writing it is used to assess. PFA, used to assess on-demand expository essays in Grades 3–12, uncovers positive and negative characteristics of a sample. To extend PFA to a new academic level and genre, I assessed scientific writing of 208 undergraduates, identifying 35 linguistic and 20 scientific prominent features. An essay could earn up to 28 positive (24 linguistic and four scientific), and up to 27 negative marks (11 linguistic and 16 scientific). The minimum prominent features number in a paper was 3, the maximum was 25 (M = 12.45, SD = 3.88). The highest positive and negative prominent features numbers noted were 17 (M = 4.11, SD = 3.96), and 16 (M = 8.34, SD = 3.25) respectively.

Rasch analysis revealed a good data-model fit, with item separation of 5.81 (.97 reliability). The estimated feature difficulty of items spanned over 10 logits; common errors were easier to avoid than “good writing” characteristics to exhibit. Significant correlations among linguistic, but not between linguistic and scientific features, suggest writing proficiency does not assure excellence in scientific writing in novices. Ten linguistic features significantly strongly and moderately inter-correlated with each other, appearing to represent writing proficiency. Student GPA correlated significantly with the raw prominent features scores (r = .37; p < .01), and negatively with the sum of negative linguistic features (r = –.40, p < .01), providing support for scale’s validity, and suggesting that good students are better at avoiding common writing errors than less able learners. Additionally, PFA scores positively significantly correlated with composite ACT scores.

To investigate PFA’s ability to track change in writing over time, I compared 2 sets of prominent features scores of 25 students. In comparison with earlier essays, later (longer) essays exhibited significantly more positive, and more negative features. Prominent features scores did not correlate significantly between the sets. This suggests, that while PFA is a valid and appropriate tool for analysis of undergraduate scientific writing, it was not suitable for tracking change in writing ability in this small sample.

Indexing (document details)
Advisor: Morse, David T.
Commitee: Elder, Anastasia D., Swain, Sherry S., Wei, Tianlan (Elaine)
School: Mississippi State University
Department: Counseling, Educational Psychology, and Special Education
School Location: United States -- Mississippi
Source: DAI-A 80/05(E), Dissertation Abstracts International
Subjects: Language arts, Educational tests & measurements, Education, Educational psychology, Science education, Higher education
Keywords: Novice scientific writing, Prominent feature analysis, Undergraduate writing, Writing assessment, Writing feedback, Writing in psychology
Publication Number: 10974642
ISBN: 978-0-438-75804-9
Copyright © 2020 ProQuest LLC. All rights reserved. Terms and Conditions Privacy Policy Cookie Policy