Structure of the Working Universe of Serials and Corrections to Faculty Score

As stated above, the ultimate purpose of the exercise with the data collected by the SRP survey was to determine whether it would be possible to bring the serials holdings of LSU Libraries up to 75% of the ST value perceived by the faculty in the "desired universe" of the 33 curriculum cores with the resources remaining in the "working universe" of these cores even after the massive cancellations and the policy of adding no new subscriptions of the late 1980s and early 1990s.

Briefly recapitulated, the "desired universe" was all those serials listed by the LSU faculty on their SRP survey forms within the 45-title limit, whereas the "working universe" consisted of the following three categories: (1) all the serials within the desired universe on current subscription at LSU Libraries; (2) the faculty score zero class of serials on current subscription at LSU Libraries, i.e., those titles classed within the 33 curriculum cores either not listed by the faculty on the survey forms or listed beyond the 45-title limit; and (3) those highly rated titles that accounted for 75% of the faculty score within the desired universe of each curriculum core but that were not on current subscription at LSU Libraries. This working universe consisted of 1,687 titles with an aggregate faculty score of 48,161 and costing $1,081,989. The mean faculty score per title was 28.5, and the mean cost per title was $641.37.

Evaluator runs were made on the working universe with the second algorithm set at the default values of trying to reduce the total costs by 75% while retaining 75% of total ST value. Given the purpose of the exercise, it was decided to place all the titles in the third category of the working universe above on subscription, even in those cases where they were no longer among the highly ranked titles accounting for 75% of faculty score due to a change in the total faculty score divisor caused by the differing serials composition of the desired and working universe cores. Moreover, the decision was made to cancel all titles in the faculty score zero class despite some of these titles not being among the highly priced ones that accounted for 75% of the total cost of their respective curriculum cores.

Due to the faults discovered in faculty ratings during their validation, it was deemed necessary to adjust both the subscription and cancellation decisions by applying citation and library use correctives to the faculty score measure. The adjustment to subscriptions necessitated a restructuring of the working universe, and it concerned the highly skewed character of informetric distributions. As has been seen above, on the average, when the serials were ranked in descending order by faculty score, more titles were required to raise the faculty score in each curriculum core another 25% from 51% to 75% than to reach the first 50%. This phenomenon resulted from the rapidly decreasing interval distances between the ordinal ranks as one went down the list. The titles that accounted for the first 50% were not considered a problem, because the level of faculty consensus was so high. However, the faculty scores of the serials in the next 25% were much smaller, and at the bottom of the range approximated the scores of those at the top of the range of the lower majority of the titles accounting for only 25% of the faculty score. Therefore the decision was made to eliminate from the working universe those serials not on subscription at LSU Libraries and located in the range from 51% to 75% of faculty score of the desired universe of the curriculum cores if they did not also meet the criterion of having an impact factor above the median impact factor of the largest of their respective ISI subject groups.Before making this adjustment, it was considered necessary to place on subscription 186 titles with an aggregate faculty score of 6,888 and a total cost of $114,066. The adjustment reduced the subscription list by 68 titles (36.6%) to 118, the added aggregate faculty score by 1,568 (22.8%) to 5,320, and the total cost of the new subscriptions by $32,184 (28.2%) to $81,882. As expected, the mean faculty score of the serials recommended for subscription rose 21.9% from 37.0 to 45.1, but their mean price also increased 13.2% from $613.26 to $693.92. The elimination of the low impact factor titles from the working universe reduced the number of titles in this universe by 4.0% to 1,619, the aggregate faculty score by 3.3% to 46,593, and the total cost by 3.0% to $1,049,805. In return, the mean faculty score per title in the adjusted working universe rose by 1.1% to 28.8, and the mean cost per title also rose by 1.1% to $648.43.

The adjustments to cancellations were regarded as necessary not only to correct for errors that resulted from the small differences in interval distances at the borderline of the highly rated titles accounting for 75% of the faculty score with the lower rated tiles accounting for only 25% of the faculty score but also to offset the complete failure in faculty perception apparent in the high NOTIS use of faculty score zero class titles. Unlike the subscription adjustments, those to cancellations did not require a restructuring of the working universe. The cancellation adjustments were twofold: (1) no title could be canceled if it had an impact factor above the median impact factor of the largest of its ISI subject groups; and (2) none of the 80 titles accounting for 80.2% of the NOTIS use of the faculty score zero class titles could be canceled. Without these adjustments, 528 titles with an aggregate faculty score of 2,336 and a total cost of $404,674 were subject to cancellation. The two adjustments reduced the number of canceled titles by 186 (35.2%) to 342, the lost aggregate faculty score by 1,002 (42.9%) to 1,334, and the cost reduction by $182,265 (45.0%) to $222,409. Surprisingly, the mean faculty score of the canceled titles fell 11.4% from 4.4 to 3.9 as a result of the adjustments. This had been totally unexpected because it had been thought that the elimination of so many titles with zero scores would raise the mean faculty score of the canceled titles, and this phenomenon again testifies to the strong correlation of faculty ratings with citation counts. However, the mean cost of the canceled titles also fell —15.1% from $766.43 to $650.32.

Tables 8 and 9 show the working universe with the new subscriptions and cancellations resulting from the Evaluator runs. As is usual with databases in library and information science, the working universe was characterized by highly skewed statistical distributions. Five (15.2%) of the curriculum cores —Biology, Chemistry, Mathematics, Physics, and Zoology & Physiology—dominated in number of titles (39.7%), faculty score (41.0%), and costs (59.3%). However, faculty score's high Spearman rank-order correlation coefficients of 0.86 with number of titles and 0.82 with costs demonstrated that the number of titles and costs of the curriculum cores were roughly in accord with their importance to LSU's ST faculty as a whole.

Interestingly enough, the five dominant curriculum cores were all derived from the Q or basic science schedules of the LC classification system and were related to academic units in the Colleges of Basic Sciences and Arts and Science. Therefore, the correspondence of the cores in terms of titles and costs to their importance can be seen not only as the result of Garfield's law of concentration, but also of the reliance of technology represented by the Colleges of Agriculture and Engineering on basic science literature.

However, what is also noticeable is that the concentration of costs on these five dominant cores was higher than the concentration of titles and faculty score. This phenomenon was the result of the high average costs of the titles in Chemistry and Physics, which alone accounted for 33.8% of the total costs of the working universe. Physics titles had the highest mean cost of $1,902.66, which was 3.8 times higher than the $504.57 mean cost for all the cores and 2.2 times higher than third-highest mean cost of $849.54 for Biochemistry. Chemistry's second-highest mean cost of $1,557.60 was 3.1 times more than the mean cost for all the cores and 1.8 times more than third-highest Biochemistry.