Academic journal article Journal of International Students

Nonresident Undergraduates' Performance in English Writing Classes-Hierarchical Linear Modeling Analysis

Academic journal article Journal of International Students

Nonresident Undergraduates' Performance in English Writing Classes-Hierarchical Linear Modeling Analysis

Article excerpt

Abstract

Do undergraduates whose native language is not English have writing deficiencies leading to academic struggles? The present study showed that the answer to this question was "no" at an American West Coast public university. This university's nonresident undergraduates on average earned B- to B+ in their colleges' English intensive-writing programs' classes, C in community college English classes, and term grade point averages between 2.5 (C+ to B-) and 3.2 (B) in the fall term of the five most recent academic years. Hierarchical linear modeling analyses showed that the predictors with the largest effect sizes were English writing programs and class level; however, each predictor accounted for less than 25% of the total variance.

Keywords: Academic success, English as a second language, international undergraduates, permanent residents, TOEFL, writing

How can American universities maximize the academic success of their nonresident undergraduates whose native language is not English? This question has become increasingly important in recent years due to the dramatic increase in the nonimmigrant international undergraduate population attending American universities (Institute for International Education [IIE], 2013a). Universities' admissions offices potentially could maximize the likelihood that admitted applicants will succeed academically by establishing appropriate entrance requirements.

One entrance requirement that many American universities have been using to predict applicants' academic success is the Test of English as a Foreign Language (TOEFL). Approximately 260 American universities require nonresident applicants whose native language is not English to submit TOEFL scores (American Exam Services, 2013); TOEFL scores are used as an indicator of English proficiency to predict future academic success (Andrade, 2006).

If English proficiency is a valid predictor of academic success for nonresident applicants who are not native English speakers (Andrade, 2006), then admitted applicants who subsequently struggle with English-despite having acceptable TOEFL scores-might be expected to struggle academically. To the contrary, a recent study (Fass-Holmes & Vaughn, 2014) demonstrated that at one American university the majority of nonimmigrant international undergraduates succeeded academically (term grade point averages [GPA] above 2.0 [C]) despite showing evidence of struggling with English. The evidence was that a majority of these students failed the university's mandatory English writing proficiency exam, and they were required to attend community college classes in English Composition and/or English as a Second Language (ESL).

Nonresident undergraduates' English proficiency and academic success could be influenced by numerous variables, some of which are student-specific and readily accessible for statistical analysis (e.g., students' citizenship country, class level, etc.), others are school-specific (e.g., classes which are taught in a particular academic term versus ones that span across several academic terms, majors within academic departments, colleges within universities, etc.), and others are unknown and/or cost-prohibitive to collect (e.g., parents' English proficiency, parents' highest level of education, etc.) (Osborne, 2000; Raudenbush & Bryk, 2002). Such variables need to be managed properly, and hierarchical linear modeling (HLM) offers many advantages in this regard. HLM analyzes data that are nested at multiple levels, computes an estimation of individual effects, partitions variance across levels, determines how much variance is accounted for at individual and group levels (Raudenbush & Bryk, 2002), uses full-information maximum likelihood estimation to handle missing data (Little & Rubin, 2002) and avoids the need to use multiple imputations (Little & Rubin, 2002). This statistical technique is more advantageous than ordinary least squares (OLS) regression, another predictive statistical technique, because OLS regression assumes independence of observations; nested data rarely fulfill this key assumption (Ker, 2014; Raudenbush & Bryk, 2002). …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.