Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
IV Computer Science Programs In this chapter 58 research-doctorate programs in computer sciences are assessed. These programs, according to the information supplied by their universities, have accounted for 1,154 doctoral degrees awarded during the FY1976-80 period--approximately 86 percent of the aggregate number of computer science and computer engineering doctorates earned from U.S. universities in this five-year span. Because computer sciences is a younger discipline than the other . . . . . . . five mathematical and physical sciences covered in this assessment and because computer science programs may be found in a variety of settings within universities, the committee encountered some difficulty in identifying research-doctorate programs that have produced graduates in this discipline. On the average, 41 full-time and part-time students intending to earn doctorates were enrolled in program in December 1980, with an average faculty size of 16 members.2 Most of the 58 programs, listed in Table 4.1, are located in computer science or computer and information science departments. Approximately 20 percent are found in departments of electrical engineering. Fifteen programs were initiated since 1970, and no two programs are located in the same university. In addition to the 58 institutions represented in this discipline, another 7 were initially identified as meeting the criteria' for inclusion in the assessment: University of Chicago George Washington University Harvard University Northeastern University data from the NRC's Survey of Earned Doctorates indicate that 889 research doctorates in computer sciences and another 458 research doctorates in computer engineering were awarded by U.S. universities between FY1976 and FY1980. 2See the reported means for measures 03 and 01 in Table 4.2. Has mentioned in Chapter I, the primary criterion for inclusion was that a university had awarded at least 5 doctorates in computer sciences during the FY1976-78 period. 59
60 Purdue University University of Southwest Louisiana University of Texas, Health Science Center--Dallas The latter two institutions chose not to participate in the assessment in any discipline. Computer science programs at the other five institutions have not bee. included in the evaluations in this discipline, since in each case the study coordinator either indicated that the institution did not at that time have a research-doctorate program in computer sciences or failed to provide the information requested by the committee. Before examining individual program results presented in Table 4.1, the reader is urged to refer to Chapter II, in which each of the 16 measures used in the assessment is discussed. Summary statistics describing every measure are given in Table 4.2. For nine of the measures, data are reported for at least 56 of the 58 computer science programs. For measures 04-07, which pertain to characteristics of the program graduates, data are presented for only approximately half of the programs; the other half had too few graduates on which to base statistics .4 For measure 12, a composite index of the size of a university library, data are available for 49 programs; for measure 14, total university expenditures for research in this discipline, data are available for 44 programs. The programs not evaluated on measures 12 and 14 are typically smaller--in terms of faculty size and graduate student enrollment--than other computer science programs. Were data on these two measures available for all 58 programs, it is likely that their reported means would be appreciably lower (and that some of the correlations of these measures with others would be higher). With respect to measure 13, the fraction of faculty with research support from the National Science Foundation, the National Institutes of Health, and the Alcohol, Drug Abuse, and Mental Health Administration, data are reported for 45 programs that had at least 10 faculty members. Intercorrelations among the 16 measures (Pearson product-moment coefficients) are given in Table 4.3. Of particular note are the high positive correlations of the measures of program size (01-03) with measures of publication records (15, 16) and reputational survey ratings (08 and 09~. Figure 4.1 illustrates the relation between the mean rating of the scholarly quality of faculty (measure 08) and the number of faculty members (measure 01) for each of 57 programs in computer sciences. Figure 4.2 plots the mean rating of program effectiveness (measure 09) against the total number of FY1976-80 program graduates (measure 021. Although in both figures there is a significant positive correlation between program size and reputa 4As mentioned in Chapter II, data for measures 04-07 are not reported if they are based on the survey responses of fewer than 10 FY1975-79 program graduates.
61 tional rating, it is quite apparent that some of the smaller programs received high mean ratings and that some of the larger programs received low mean ratings. Table 4.4 describes the 108 faculty members who participated in the evaluation of computer science programs. These individuals con- stituted 62 percent of those asked to respond to the survey in this discipline and 12 percent of the faculty population in the 58 research-doctorate programs being evaluated.S A majority of the survey participants had earned their highest degree since 1970, and almost one-third held the rank of assistant professor. Two exception should be noted with regard to the survey evaluations in this disci- pline. Regretably, ratings are unavailable for the program in the Department of Computer and Communications Sciences at the University of Michigan since an entirely inaccurate list of its faculty members was included on the survey form. Also, it has been called to the attention of the committee that the faculty list (used in the survey) for the Department of Computer Science at Columbia University was missing the names of four members. The committee has decided to report the survey results for this program but cautions that the reputational ratings may have been influenced by the omission of these names. To assist the reader in interpreting results of the survey evaluations, estimated standard errors have been computed for mean ratings of the scholarly quality of faculty in 57 computer science programs (and are given in Table 4.1~. For each program the mean rating and an associated "confidence interval" of 1.5 standard errors are illustrated in Figure 4.3 (listed in order of highest to lowest mean rating). In comparing two programs, if their confidence intervals do not overlap, one may conclude that there is a significant difference in their mean ratings at a .05 level of significance.6 From this figure it is also apparent that one should have somewhat more confidence in the accuracy of the mean ratings of higher-rated programs than lower-rated programs. This generalization results primarily from the fact that evaluators are not as likely to be familiar with the less prestigious programs, and consequently the mean ratings of these programs are usually based on fewer survey responses. 5 See Table 2.3 in Chapter II. 6 See pp. 29-31 for a discussion of the interpretation of mean ratings and associated confidence intervals. s
62 TABLE 4.1 Program Measures (Raw and Standardized Values) in Computer Sciences Prog No. University - Department/Academic Unit 001. Arizona, University of-Tucson Computer Sciences 002. Brown University Computer Science* 003. California Institute of Technology Computer Science* 004. California, University of-Berkeley Electrical Engineering & Computer Sciences 005. California, University of-Irvine Information and Computer Science 006. California, University of-Los Angeles Computer Science 007. California, University of-San Diego Electrical Engineering & Computer Science 008. California, University of-Santa Barbara Electrical and Computer Engineering 009. Carnegie-Mellon University Computer Science 010. Case Western Reserve University Computer Engin/Computing & Information Sci 011. Columbia University Computer Science* 012. Connecticut, University of-Storrs Electrical Engineering & Computer Science 013. Cornell University-Ithaca Computer Science 014. Duke University Computer Science* Georgia Institute of Technology Information and Computer Science 016. Illinois, University-Urbana/Champaign Computer Science 017. Indiana University-Bloomington Computer Science* 018. Iowa State University-Ames Computer Science 019. Iowa, University of-Iowa City Computer Science 020. Kansas State University-Manhattan Computer Science* * indicates program was initiated since 1970. Program Size (01) (02) (03) 7 4 40 42 8 4 42 42 5 5 37 42 30 43 66 62 12 15 45 47 36 55 73 68 9 7 42 43 8 10 41 45 31 41 68 61 7 9 40 44 11 2 44 41 11 8 44 44 14 34 48 57 13 10 47 45 11 8 44 44 30 112 66 97 15 NA 49 15 18 49 49 12 11 45 45 8 7 41 43 13 43 21 45 22 45 53 53 46 51 103 65 17 44 9 42 83 60 10 42 17 44 12 43 48 .27 52 58 29 NA 47 30 47 125 71 16 44 17 .07 44 41 20 .36 45 67 12 NA 43 Characteristics of Program Graduates (04) (05) (06) (07) NA NA NA NA NA NA 14 6.3 47 51 09 5.3 43 60 00 7.9 35 37 NA NA NA NA 22 6.8 54 47 NA NA NA NA NA NA 10 NA 43 13 6.1 46 53 NA NA NA NA NA .79 50 .64 32 .65 34 NA NA .81 51 NA NA NA 5.5 .97 59 69 NA NA .80 50 .85 56 NA 5.8 .73 56 43 5.3 .80 61 50 NA NA NOTE: On the first line of data for every program, raw values for each measure are reported; on the second line values are reported in standardized form, with mean - 50 and standard deviation 3 LO- ANAL indicates that the value for a measure is not available. NA NA NA 24 41 27 43 25 42 NA NA 50 57 NA NA NA 63 65 NA 10 33 28 44 NA 33 47 70 69 NA
63 TABLE 4.1 Program Measures (Raw and Standardized Values) in Computer Sciences University Research Published Survey Ratings Prog Survev Results Librarv Support Articles Standard Error No. (08) (09) (10) (11) (12) (13) (14) (15)(16) (08) (09) (10) (11) 001. 2.4 1.4 1.3 1.0 0.9 NA 323 15 .11 .10 .11 .07 48 48 57 54 55 44 4446 002. 2.9 1.7 1.5 1.2 - 1.1 NA 417 25 .10 .09 .08 .06 54 53 63 57 35 45 4747 003. 2.5 1.5 0.8 0.9 NA NA 871 26 .17 .11 .12 .08 50 50 40 50 48 4847 004. 4.5 2.6 1.3 1.6 2.2 .60 NA 134 .08 .06 .07 .05 70 69 57 68 69 63 8382 005. 2.4 1.4 0.9 0.8 NA .17 98 12 .09 .09 .13 .07 49 47 42 50 40 43 4344 006. 3.8 2.2 1.3 1.3 2.0 .61 126 77 .08 .05 .08 .05 63 62 5 7 60 66 63 43 6461 007. 2.6 1.2 1.1 0.8 -0.0 NA 376 21 .13 .12 .11 .07 51 45 49 49 45 45 4648 008. 2.1 1.2 1.1 0.8 -0.1 NA 305 27 .11 .11 .12 .07 46 43 50 48 45 44 4849 009. 4.8 2.7 1.1 1.8 NA .26 3649 53 .05 .05 .07 .05 73 71 50 72 44 6 7 5661 010. 1.3 0.8 0.4 0.4 -1.3 NA NA 24 .11 .13 .10 .06 37 36 24 40 32 4744 011. 2.5 1.2 1.6 0.8 1.7 .36 NA 23 .12 .11 .08 .07 50 45 67 49 64 50 4746 012. 1.7 1.0 1.2 0.5 -0.5 .36 435 12 .13 .12 .09 .06 41 41 54 43 41 50 45 4343 013. 4.3 2.5 1.1 1.6 1.6 .57 987 52 .07 .06 .07 .05 68 68 49 67 62 61 49 5654 014. 2.4 1.5 1.3 0.7 0.3 .46 218 12 .10 .10 .10 .07 49 50 56 46 49 55 44 4343 015. 2.7 1.6 1.8 0.8 NA .27 4056 30 .10 .08 .06 .07 52 51 75 50 45 69 4948 016. 3.8 2.3 1.0 1.4 2.0 .53 3357 155 .09 .07 .07 .06 63 63 46 62 66 59 65 8983 017. 2.3 1.3 1.6 0.8 0.9 .53 67 20 .11 .12 .08 .07 48 46 66 49 55 59 43 4646 018. 1.7 1.2 1.2 0.4 - 0.5 .27 NA 14 .12 .13 .09 .06 42 45 51 40 40 45 4443 019. 1.7 1.1 1.0 0.5 0.3 .25 355 10 .12 .10 .11 .07 41 43 44 41 49 44 45 4242 020. 0.9 0.6 0.9 0.2 NA NA 153 13 .13 .11 .10 .04 33 33 41 34 43 4342 NOTE: On the first line of data for every program, raw values for each measure are reported; on the second line values are reported in standardized form, with mean a 50 and standard deviation ~ 10. GNAT indicates that the value for a measure is not available. Since the scale used to compute measure (16) is entirely arbitrary, only values in standardized form are reported for this measure.
64 TABLE 4.1 Program Measures (Raw and Standardized Values) in Computer Sciences Characteristics of Prog Program Size Program Graduates No. University - Department/Academic Unit (01) (02) (03) (04) (05) (06) (07) 021. Kansas, University of 16 6 6 NA NA NA NA Computer Science* 50 43 41 022. Maryland, University of-College Park 28 35 50 .08 8.0 .76 .24 Computer Science 64 58 52 42 37 46 41 023. Massachusetts Institute of Technology 34 62 135 .23 6.4 .77 .39 El ectrical Engineering and Computer Science 71 71 73 55 51 47 50 024. Massachusetts, University of-Amherst 16 16 70 .00 5.4 .82 .36 Computer and Information Sciences* 50 48 57 35 S9 52 49 025. Michigan State University-East Lansing 14 8 20 NA NA NA NA Computer Science 48 44 45 026. Michigan, University of-Ann Arbor 10 18 38 .30 6.4 .79 .63 Computer and Communication Sciences 43 49 49 61 50 49 65 027. Minnesota, University of 21 13 28 .25 6.5 .81 .50 Computer Science 56 46 47 57 50 52 57 028. Missouri, University of-Rolla 11 9 11 NA NA NA NA Computer Science* 44 4.4 42 029. New York University 13 19 63 .21 8.0 .80 .20 Computer Science 47 49 55 53 37 50 39 030. North Carolina, University of-Chapel Hill 8 17 23 .07 7.5 .88 .19 Computer Science 41 48 45 41 41 59 38 031. Northwestern University 24 41 16 .08 6.5 .92 .33 Electrical Engineering & Computer Sciences 59 61 44 41 50 63 47 032. Ohio State University-Columbus 21 43 90 .09 6.3 .76 .48 Computer and Information Science 56 62 62 43 51 46 56 033. Oklahoma, University of-Norman 15 1 26 NA NA NA NA Electrical Engineering and Computer Sci 49 40 46 034. Pennsylvania State University 16 10 32 .15 6.3 .77 .39 Computer Sciences 50 45 48 48 52 47 50 035. Pennsylvania, University of 29 25 54 .14 6.4 .79 .41 Computer and In formation Science 65 52 53 47 51 50 52 036. Pittsburgh, University of 12 10 20 NA NA NA NA Computer Science* 45 45 45 037. Polytech Institute of New York 8 6 33 NA NA NA NA Electrical Engineering and Computer Science 41 43 48 038. Princeton University 9 21 26 .23 4.3 1.00 .60 El ectrical Engineering and Computer Science 42 50 46 55 69 72 63 039. Rice University 22 19 23 .29 4.9 .52 .14 Mathematical Sciences 5 7 49 45 60 63 20 35 040. Rochester, University of 11 5 36 NA NA NA NA Computer Science* 44 42 49 * indicates program was initiated since 1970. NOTE: On the first line of data for every program, raw values for each measure are reported; on the second line values are reported in standardized form, with mean = 50 and standard deviation = 10. "NA" indicates that the value for a measure is not available.
69 TABLE 4.3 Intercorrelations Among Program Measures on 58 Programs in Computer Sciences Measure 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 Program Size 01 .62 .67 -.11 -.03-.17-.26 .54 .54 .13 .45.28 .12.44.62.61 02 .52 .05 -.07.12-.05 .66 .68 -.02 .61.44 .34.59.85.84 03 .04- .05 - .06 - .07 .50 .49 .12 .4 6 .33 .13 .43 .51 . 52 Program Graduates 04 .12 .17 .43 .35 .34 -.07 .30 .23 .34 .22 .09 .20 05 .13.44 .14 .17 -.22 .10 -.17 .29 -.21 -.07 -.04 06 .41 .21 .26 -.11 .25 .23 .26 -.03 .10 .14 07 .17 .23 -.31 .23 .24 .26 -.16 -.08 -.01 Survey Results 08 .98.29 .97 .58 .59 .63 .70 .77 09 .26 .95 .54 .61 .61 .69 .75 10 .26 .16 .18 -.02 .04 .05 11 .56 .57 .64 .69 . 74 University Library 12 .49 .16 .52 .52 Research Suppor t 13 14 Publication Records 15 16 .10 .32 .35 .66 .73 .98 NOTE: Since in computing correlation coefficients program data must be available for both of the measures being correlated, the actual number of programs on which each coefficient is based varies.
70 s ~ o++ + 4.0++ Measure + 3.0++ 08 + 2.0++ * * * * * * * * * + * 1 + * * * * + * + * * + * * * * * * * ' O+ * * * * * * * * * * * * 0.0 +/+++++++++++++++/+++++++++++++++/+++++++++++++++/+++++++++++++++/+++++++++++++++/+++++++++++++++/ 1 4 9 16 25 36 49 Measure 01 (square root ~cale) FIGURE 4.1 Mean rating of ~cholarly quality of faculty (measure 08) versus number of faculty members (measure 01)--57 programs in computer sciences.
71 3 . 0++ + + 2. 0++ + + Messure + * * * * 09 + *. * * + * 1. 0++ + + + O.O +/+++++++++/+++++++++/+++++++++/+++++++++/+++++++++/+++++++++/+++++++++/+++++++++/+++++++++/+++++++++/ 1 4 9 16 25 36 49 64 81 100 121 * * * * * * * * * * * *, * * * * * * * ** * * * * * * * * * : ** * * * * * * r ~ .71 Measure 02 (square root scale) FIGURE 4.2 Mean rating of program effectiveness in educating research scholars/scientists (measure 09) versus number of graduates in last f ive years (measure 02)--56 programs in computer sciences.
72 TABLE 4.4 Characteristics of Survey Participants in Computer Sciences Respondents N% Field of Specialization Computer Sciences9992 Other/Unknown98 Faculty Rank Professor4138 Associate Professor3230 Assistant Professor3432 Other/Unknown11 Year of Highest Degree Pre-195033 1950-591110 1960-693028 Post-19696358 Unknown11 Evaluator Selection Nominated by Institution8175 Other2725 Survey Form With Faculty Names9790 Without Names1110 Total Evaluators108100 l
73 X- x 4.0 3.0 2.0 Mean Survey Rating (treasure 08) FIGURE 4.3 Mean rating of scholarly quality of faculty in 57 programs in computer sciences. 1.0 o.o NOTE: Programs are listed in sequence of mean rating, with the highest-rated program appearing at the top of the page. The broken lines (---) indicate a confidence interval of +1.5 standard errors around the reported mean (x) of each program.
