DOCUMENT RESUME ED 345 781 JC 920 261 AUTHOR Isonio, Steven TITLE Implementation and Initial Validation of the APS and English Test The APS English-Writing Test at Golden West College: Evidence for Predictive Validity. INSTITUTION Golden West Coll., Huntington Beach, Calif. PUB DATE 91 NCTE 39p. PUB TYPE Reports - Evaluative/Feasibility (142) -- Statistical Data (110) RDRS PRICE XF01/PCO2 Plus Postage. DESCRIPTORS *College English; Community Colleges; Correlation; Grades (Scholastic); Predictive Neasurement; Predictive Validity; Predictor Variables; *Reading Tests; 'Student Evaluation; 'Student Placement; Tables (Data); Test Results; Test Use; *Test Validity; Two Year Colleges; Writing (Composition); 'Writing Tests IDENTIFIERS Assessment and Placement Services for Comm Coll ABSTRACT In May 1991, Golden West College (California) conducted a validation study of the English portion of the Assessment and Placement Services for Community Colleges (APS), followed by a predictive validity study in July 1991. The initial study was designed to aid in the implementation of the new test at GWC by comparing data on APS use at other community colleges and by testing placement rules. The follow-up study examined the relationship between the APS and eventual course success as measured by students' grades. The initial sample for both studies consisted of 381 students from 23 classes offered in spring 1991 who had taken the reading and writing portions of the APS. Three standard criterton groups were identified, consisting of native English speakers placed in English 009, 010, or 100. Selected findings Included the iollowing: (1) the APS appeared able to differentiate among course levels and was significantly related to standard group membership, making a clear placement recommendation possible for the majority of students; (2) ratings of writing samples could aid in definitive placement of those students earning unclear APS scores; (3) the gain in predictability from using a combination of reading and writing score information over using writing alone was insignificaLt; (4) the analysis of the relationship of grades to APS sc6res revealed a low to moderate-sized predictive validity coefficient, a value within the expected range; and (5) a significant difference in APS scores was found between successful and unsuccessful students. Data tables, references and short appendixes, providing a sample student survey and a correction for restricted range of a predictive validity coefficient, are included. (JSP) *******************0**************************0************************ Reproductions suppliee by EDRS are the Dest that can be made from the original document. ***************************************************************ft******* 1 Implementation and Initial Validation of the APS English Test and The APS English-Writing Test at Golden West College: Evidence for Predictive Validity GOLDEN WEST COLLEGE 1991 S DEPARTMENT CIF EDUCATION U ottre of Eaucahonat Research srid omprowemeni "PERMISSIONTOREPRODUCETHIS EDUCATIONAL RESOURCES ISWORMATION MATERIAL HAS BEEN GRANTED BY CENTER (ERIC{ Min aecumaM has been reproduced an S. Isonio the pirson or organization received from hare Ivan rnatie to improve Minor changes optocfacfron QUSIIIy (loco Points i.vf view or opimonii %fated in tn mant do not necessarriv represent ofirrize TO THE EDUCATIONAL RESOURCES OERI positron or policy INFORMATION CENTER (ERIC).- Steven Isonio, PhD Advisor, Matriculation Research/Assessment Cqrv pc!Tn BTST L 2 i I a Implementation and Initial Validation of the APS English Test GOLDEN WEST COLLEGE May 1991 Steven bonito, PhD Advisor, Matriculation Research/Assessment Validation of the Implementation and Initial APS English Test background, is usLd to present the A question and answer format assessment validation study procedures and results of an the study was The primary intention of conducted in Spring 2991. West for the new test at Golden to establish placement rules followup research is An overview of necessary College. presented. implementation of iu the adoption and What steps are involved I. into writing skills for placing students a new measure of College? English courses at Goldet West recognizing the need for a better The first steps, namely, consideration, discussing its test, identifying a test for A adopting it, are complete. merits, and finally formally portion of the Assessment and proposal to adopt the English Colleges (APS), developed and Placement Services for Community Service (ETS) was made in distributed by the Educational Testing recommended by a The test was reviewed and December 1990. In February 1991, department. subcommittee of the GWC English faculty voted 14-0 (with one the English Department full-time abstention) to adopt the new test. would designed to gather information that A pilot study was placement rules as well as begin help with the establishment of and the relationship between the APS the process of documenting research is Such assessment validation relevant criteria. likely, Adjustments in placement rules are necessarily ongoing. studies are required. and an asssortment of validation Stanford TASK for placing Is the APS a better test than the 2. students into courses at Golden West? validity, there is a consensus While both tests have face the of higher level skills than that the APS taps a broader range Academic Skills), thereby making it more TASK (Stanford Test of Both into writing courses. appropriate for placing students the English department instruments were thoroughly reviewed by This review resulted in a strong assessment subcommittee. TASK as the objective recommendation that the APS replace the The TASK was English placement test at Golden West. discrete items" characterized as "simplistic, concentrating on Conversely, the APS was ability". and evaluating "a low level of it "features a higher level of deemed more appropriate in that global scope." writing and language skills and has a more developed specifjcally Further, unlike the TASK, the APS was skills' college students into basic for placing of community widely used Additionally, the APS is now the most courses. into English courses in instrument for placing students California Community Colleges. 1 What guidelines for use of the APS 3. are offered by the test publisher? ETS provides extensive information about the construction and norming of the test in tha technical manual (College Entrance Examination Board, 1990). References to pertinent research are also provided. Additionally, the manual contains a discussion of selected psychometric properties of the test. However, this instrument, like any placement test, does not come with universally applicable placemert rules. Local placement rules must be developed. What type of placement rules 4. are used at other schools that have adopted the APS? Information about how the instrument is used at a sample of other colleges was gathered. This information suggested models that would be tested as a part of the pilot Table 1 study. presents placement rules implemented at a sample of four colleges that use the APS for placing students into three levels of English. To facilitate comparison, course numbers for the transferable composition course have been changed to "100"; similarly, the two courses below this level are referred to as "10" and "9", respectively, although these may not be the actual course numbers at the other colleges. Two schools, American River College and Rancho Santiago College use only the objective writing section of the APS to place students W.rectly into the three courses. Moorpark College and Santa Barbara City College incorporate information from an essay to clarify placements in certain marginal ranges. Specifically, in the case of Moorpark College, students who scored in the upper portion of the range for placement into the English 10 course can challenge this placement by having their essay evaluated. The student's placement is either changed to the English 100 course, or remains English 10. Santa Barbara City College routinely evaluates essays written by students in either of three ranges: those students in a band between English 10 and English 100, those students in a band between English 9 and English 10, and those students scoring in the very low ranges of the distribution. There appears to be much consistency in placement rules used by these schools. This information was useful in helping to generate reasonable placement rules for consideration in the study. 2 1 Table Rules Three-tiered Placement Colleges at Sample With APS Writing Test RANCHO Score SANTIAGO SANTA BARBARA AM.RIV. MOORPARX 100 100 100 100 40 100 100 100 100 39 100 100 100 100 38 100 100 100 100 37 100 100 100 100 36 100 100 100 100 36 100 100 100 100 35 100 100 100 100 33 100 100 100 100 32 100 10 OR 100 100 100 31 100 10 OR 100 100 100 30 10 10 OR 100 100 100 29 10 10 100 100 28 10 10 10 10+ 27 10 10 10 10+ 26 10 10 10 10+ 25 10 10 10 10+ 24 9 10 10 10 23 10 OR 9 9 10 10 22 9 10 OR 9 10 10 21 9 10 OR 9 9 10 20 10 OR 9 9 9 10 19 9 9 9 10 28 9 9 10 9 17 9 9 9 9 16 9 9 9 9 15 9 9 OR LOWER 9 9 14 9 9 OR LOWER 9 9 13 9 9 OR LOWER 9 9 12 9 9 OR LOWER 9 9 11 9 9 OR LOWER 9 9 10 9 9 OR LOWER 9 9 9 9 9 OR LOWER 9 9 8 9 9 OR LOWER 9 9 7 9 9 OR LOWER 9 9 6 9 9 OR LOWER 9 9 5 9 9 OR LOWER 9 9 4 9 9 OR LOWER 9 9 3 9 9 OR LOWER 9 9 2 9 9 OR LOWER 9 9 3, indicated, definitive two placements are In cases where rote. "10+11 of the essay. an evaluation placement is made based upon students level course, but the English 10 indicates placement into placement into read for possible that their essay be may request the English 100 course. Describe the design and procedures of the pilot study. what 5. was expected? In the Spring 1991 semester, a representative sample of 23 A total of 381 English 9, 10, and 100 classes was identifed. students from these classes were administered the reading and Additionally, writing sections of the APS (see Table 2). information about the students' placement based upon the TASK test (previously used objective English placement test), a wrfting sample produced the first day of instruction, and instructors' recommendations was gathered. These Three "standard" criterion groups were identified. groups consisted of native English speakers who were placed into English 9, 10, or 100 and for whom these indicators were in Certainly this composite criterion is not perfect, agreement. but its reliability was enhanced by the convergence of these Finally, only native speakers of English various indicators. were included in the analysis because the APS was designed Primary language specifically for placement of such students. was determined from a supplemental survey administered after the test (see Appendix A). Distributions of scores were obtained for the reading and The relationship writing sections of APS, and their combination. between APS scores and course level was determined with the To the multiserial correlation coefficient (Jaspen, 1946). extent that a relationship exists between the APs scores and standard group membership, the APS can be said to have criterion- Further, the central points of the three related validity. distributions should be at different locations along the abscissa, ideally with a only a moderate degree of overlap. Finally, a variety of placement rules were evaluated. The ability of these various rules to differentiate among the three groups (maximizing correct "placements" while minimizing The score ranges for the most incorrect ones) was assessed. effective model would constitute the recommended initial cut scores. Is the APS capable of differentiating the three levels of 6. English courses? The process of implementing a test and establishing placement rules entails comparing it to a standard. The difficulty arises in that there is no ideal criterion to serve as Therefore, as noted above, existing information the standard. (placement based on the previous English placement test (TASK), ratings of the first day placement writing samples (PWS), and students'. self-evaluation implicit in their course selection were used to de'fine three standard comparison groups. At a minimum, APS mean scores for the three standard grpups The standard groups should be clearly different from each other. do indeed differ significantly in terms of mean hPS writing score Figure la depicts these group [f (2,187) = 79.09, p < .0013. The reading and writing scores were combined mean differences. 4 7 Table 2 Course Level Number of Students Tested by English Students in Students Number of Course sis* a the es Cla es ve 44 54 English 9 4 60 156 English 10 8 87 171 English 100 11 191 381 Total 23 placed into their *Only Native Speakers of English who were Placement Writing enrolled class with the TASK-EPT and the These students Sample were included in the analysis. comprise the standard criterion groups. 5 Figure la Mean APS Writing Score By Level 30 25 - English 100 English 010 English 009 Standard Group Level API Pam, Ipilpg URI APS Writing Scores Descriptive Statistics Standard Level - 27 5.13 7 17.82 009 English - 33 12 4.87 22.92 010 English - 38 15 4.75 28.87 100 English (Schmidt, 1971). The values for each student by summing the two the three significantly across also differed mean combined score These means are 94.50, R < .001). groups LE (2,185) = writing the mean APS Thus, for both presented in Figure lb. reading score, there are combination with the score alone and in differences amonc the three groups. very striking such as those just Differences among the group means points of the only that the central described, however, indicate to Therefore, it was necessary differ. three distributions determine the extent of scores to examine the full distributions Figures 2a, 2b, and 2c distributions overlap. to which the 010, distributions for English 009, depict the APS writing score to reflect The ordinate was scaled and 100, respectively. across the facilitate direct comparisons relative frequency to among The striking differences samples. three different-sized reflected by the writing scores noted above, are group mean sample, For the English 009 distributions. differencen among the were 12 to 33 to 27, whereas the ranges the scores ranged from 7 .21lko 38 for English 100. for English 10, and semi-interquartile also reflected in the The differences are These ranges are 15 middle 50% of the distribution. ranges, the and 27 to 32 for to 27 for English 10, to 21 for English 9, 20 semi-interquartile range is The advantage of the English 100. excluding "core" of the distribution, that it reflects the low motivation be attributable to very extreme scores which may particularly stable. Therefore, these values are or luck. of depict the distributions Similarly, Figures 3a, 3b, 3c standard for students in the combined reading and writing scores, The central points respectively. English 9, 10 and 100 groups? different places on clearly located at of the distributions are to be slightly more However, there appears axis. the APS score The levels. distributions for adjacent course overlap between For English 10 English 9 is 25-38. semi-interguartile range for 48-61. English 100 the range is this value is 33-50; for somewhat writing section alone appears In short, the APS The English. the three levels of able to differentiate among slightly less reading scores seems combination of writing and able to do so. in conjunction with be best utilized Row can a writing sample 7. maximizing accuracy of information (that is, APS test score efficiency)? placement while maintaining the sample APS is no consensus among As notad earlier, there Some schools do writing samples. schools regarding the use of to supplement Other schools use them not use writing samples. placement typically to make definitive the APS scores, in marginal ranges. recommendations for students course to differentiate among Since the APS appears able membership, related to standard group levels and is significantly for the majority of recommendation is possible a clear placement contribute to a Ratings of writing samples can students. that students who earn APS scores definitive placement for those 7 AVAILABLE dEST COPY