College and Research Libraries A. K. JAIN Sampling and Short-Period Usage in the Purdue Library Several possible methods of sampling the social science monograph titles in the general library of Purdue University were considered, and a "goo(r method was used to obtain estimates of their usage in the library and at home during the period July 1-August 4, 1964. The term relative usage was defined and used to study the effect of: (1) lan- guage, (2) country of publication, (3) year of publication, and (4) year of accession of a monograph title. An attempt was made to fit a re- gression model for titles in English by quantifying the last three in- dependent variables with relative usage as the dependent variable. Functions based on the above variables have been developed to identify monograph titles for storage. A questionnaire was employed to study the usage of library facili- ties and to gather opinions of library patrons. Purpose of visiting the library, reason for checkout of library material, reason for preferring library or home for the use of library material, etc., were analyzed on the basis of the replies received. THE usuAL APPROACH t9 studying -us- age in a library is to start with a repre- sentative sample of titles, to record us- age of these titles from their book cards (assuming such records are available) and to analyze these data. The study by Fussier and Simon1 is an excellent ex- ample. Trueswell2 has studied the ~'last circulation date" fo! books by saving the 1 Herman H. Fussier and J. L. Simon, Patterns in the Use of Books in Large Research Libraries (Chicago: University of Chicago Library, 1961). 2 Richard W. Trueswell, "A Quantitative Measure of User Circulation Requirements and Its Possible Effect on Stack Thinning and Multiple Copy De- termination," American Documentation, XVI (Janu- ary 1965). Mr. Jain is in the School of Industrial Engineering at Purdue University. This paper was read at the 1965 annual meeting of the American Society for Engineering Education in Chicago. book cards for each day's circulation and has derived some useful rules for thinning the stacks. A question remain- ing unanswered is what a librarian who · does not have book cards does to study usage of books? This paper presents an approach to studying usage when there are no book cards. Also, a librarian who has book cards may prefer to use this approach in place of the usual approach because of its simplicity and conveni- ence. This paper also presents some results of a questionnaire survey carried out to study the usage of the Purdue library facilities and to gather opinions of li- brary patrons. THE APPROACH Three independent samples of mono- graph titles from the total collection I 211 212 I College & Research Libraries • May, 1966 (S), home-used material (H), and in- library used material ( I) respectively are obtained. These samples are divided into a certain number of groups on the basis of the following characteristics of the title: ( 1) language, ( 2) country of publication, ( 3) year of publication, and ( 4) year of accession. The frequen- cies of 'titles in these groups are com- puted for each of the three samples. Let si = number of titles in sample s which belong to the ith group; H 1 = number of titles in sample H which belong to the ith group; I 1 = number of titles in sam- ple I which belong to the ith group. Then, relative usage R is defined as fol- lows: R1 = % relative usage of ith group = Ht + T t (100). St The magnitude of the indices Rt de- pends on the relative sizes of the sam- ples S, H, and I. For example, if we take a very small sample of total collection ( S ) but take all titles used at home ( H) and in-library (I) during a long period of time ( say five years) as the other two samples, then indices Rt will be large in magnitude. But, in any case, R1/RJ is independent of all possible vari- ations in samples S, H, and I, for all i and j, as long as ~H~/~It • • is fixed. Because of this, it is very useful to compare these indices R1 among themselves and the groups having the lowest Rt are the ones which are the least important from the point of view of usage. We will first discuss the method of sampling and then discuss some results based on the above indices Rt. SAMPLING ToTAL CoLLECTION, HoME UsE, AND IN-LmRARY USE This work was done during the sum- mer session (June 15-August 7) 1964. Since time was short, it was decided to confine the scope of this study to mono- graph titles in Dewey Decimal Classifi- cation ( DDC) 330-379. A "title," in- cluding all copies of all volumes of all editions on the third floor of the Purdue general library, was considered as the best sampling unit. The sampling schemes used to obtain the three sam- ples S, H, and I are discussed below. Total Collection (S). The following sampling frames were considered: ( 1 ) books on shelves on the floor; ( 2) au- thor catalog; ( 3) subject and title cata- log; and ( 4) shelf list. The first frame was rejected because all books can not be found on shelves at any point of time (our study has revealed that only 65 per cent of titles were on shelves ) and because it was not handy. The sec- ond frame was considered good since it was easy to get a representative sample of the titles ( assign weights inversely proportional to the number of authors, i.e. number of cards in author catalog). For this study, however, it would have been very wasteful because there was interest in DDC 330-379 only. The third frame sounded good. But it turned out that there were many titles and subtitles for each subject and it would have been extremely difficult to get a representa- tive sample of DDC 330-379. The last frame, viz. shelHist, was con- sidered the best. The only drawback of the shelHist frame was that the num- ber of cards for a title was not necessari- ly one (in some cases there were more than five cards for a title). This draw- back was remedied by ignoring all cards except the first one for each title. There were fifty-one drawers of shelflist cards for DDC 330-379. Since the time was short, it was decided to obtain a 5 per cent (or 1 in 20) sample of titles. The next question was what sampling method to use, systematic or random? Systematic sampling was chosen because Fussier and Simon tested a subject area for cyclical effects in systematic sam- pling and found none and because ran- Sampling and Short-Period Usage in the Purdue Library I 213 TABLE 1 LIBRARY's HoLDINGS AND RELATIVE UsAGE BY YEAR oF PuBLICATION NUMBER S IN DDC YEAR OF 330- 340- 370- PUBLICATION 339 369 379 ------ Pre-1904 14 12 13 1904-1913 28 18 27 1914-1923 36 17 41 1924-1933 45 30 42 1934-1943 73 62 58 1944-1953 105 82 81 1954-1964 139 133 169 Not Available 3 1 6 Total 443 355 437 dom sampling would have been very expensive. To get a 5 per cent sample nineteen titles had to be skipped be- tween every two titles in the sample. It would have been time-consuming to skip nineteen titles, and there was a good possibility of error in count, since cards other than the first were to be ignored. Mter considering various pos- sibilities the method used was the fol- lowing: each ' of the fifty-one drawers was divided into four equal parts (i.e. four equal lengths of cards), and one out of the four parts was selected at random. In the selected part, a random start was made by picking up one of the first five cards at random including every fifth title (systematically). · Home Use (H). There are no book cards in Purdue library. A checkout slip is filled in at the time of checkout of a title, and it is discharged after the title is returned to the library. Checkout slips for titles in DDC 330-379 returned to the library during July 1-August 4, 1964 were saved, and these titles constituted sample H. This type of sample is differ- ent from a sample consisting of titles checked out during a certain period. Both of these are subject to a bias be- cause of variations in loan periods for different classes of users. The former was chosen because it was easy to han- dle discharged slips. PER CENT s IN DDC PER CENT R IN DDC 330- 340- 370- 330- 340- 37{}- 339 369 379 339 369 379 ------------------ 3.2 3.4 3.0 43 42 231 6.3 5.1 6.2 46 22 274 8.1 4.8 9.4 39 24 85 10.1 8.4 9.6 80 43 148 16.5 17.4 13.2 86 44 238 23.7 23.1 18.5 118 77 347 31.4 37.5 38.7 165 135 598 0.7 0.3 1.4 67 0 67 100.0 100.0 100.0 110 83 374 In-Library Use (I). Library patrons pick up titles from the open shelves of the library and leave them on tables after their use. Though not very realistic, the simplifying assumption was made that all titles left on tables had been used (i.e. the fact was ignored that some of the titles left on tables might have been "rejected" instead of "used" by the library patrons). All titles in DDC 330-379 found on tables early in the morning and late in the evening during July 2- August 4, 1964 constituted sam- ple I. This sample accounted for about 67 per cent of the total material ( DDC 330-379) reshelved during the survey period. REsULTS BASED oN SHORT-PERIOD UsAGE The over-all relative usages of mono- graph titles in DDC 330-339, 340-369, and 370-379 were 110, 83, and 37 4 re- spectively. Education (i.e. DDC 370- 379) . titles had three to four times as much usage as titles in DDC 330-339 and DDC 340-369. This was mainly because many high school teachers took courses in education at Purdue during the summer session. The pattern is ex- pected to be quite different during the regular semesters. Use and Age of Title. Table 1 presents the Purdue library's holdings and rela- tive usage by year of publication. It 214 I College & Research Libraries • May, 1966 TABLE 2 LmRARY's HoLDINGS AND RELATIVE UsAGE BY YEAR OF AcCESsiON NUMBER S IN DDC YEAR OF 330- 340- 370- ACCESSION 339 369 379 --- Pre-1924 52 29 45 1924-1933 33 25 41 1934-1943 75 50 69 1944-1953 80 69 73 1954-1964 197 177 197 Not Available 6 5 12 Total 443 355 437 will be seen that, generally speaking, relative usage decreased monotonically with age (i.e. the number of years since the date of publication) for DDC 330- 339 and DDC 340-369. But for DDC 370-379 the pattern was quite different: titles published before 1904, during 1904-1913, and 1934-1943 had about the same relative usage which was higher than the relative usage for titles pub- lished during 1914-1933. Table 2 shows that the pattern by year of accession was similar to that by year of publica- tion. Use of Foreign Books. From Table 3 it is clear that the relative usage was higher for titles published in the USA and England than for those published in France, Germany, and other countries. It may be pointed out that in the Pur- due library about 85 per cent of titles in DDC 330-339 and DDC 340-369 and PER CENT s IN DDC PER CENT R IN DDC 330- 340- 370- 330- 340- 37'0- 339 369 379 339 369 379 ------------------ 11.7 8.2 10.3 23 14 184 7.4 7.0 9.4 94 24 83 16.9 14.1 15.8 77 40 203 18.1 19.4 16.7 110 61 203 44.5 49.9 45.1 148 126 621 1.4 1.4 2.7 117 0 42 100.0 100.0 100.0 110 83 374 94 per cent of titles in DDC 370-379 were published in the USA. Table 4 shows that the relative usage was the highest for titles in English for each of the three groups. The next im- portant language from the point of view of usage was French. It is interesting to note that all used titles (i.e. samples H and I) in DDC 340-369 were in English. It must be mentioned that since 94 per cent or more of the titles in the Purdue library in DDC 330-379 are in English, it is necessary to use larger sample sizes before any definite conclu- sions can be drawn regarding relative usage of titles in languages other than English. Usage by Class of User. The Table 5 shows the home usage of monograph titles in D DC 330-379 by class of user. It is clear that graduate students used TABLE 3 LmRARY's HoLDINGS AND RELATIVE UsAGE BY CoUNTRY OF PUBLICATION NUMBER S IN DDC PER CENT s IN DDC PER CENT R IN DDC 330- 840- 370- 330- 340- 370- 330- 340- 37'0- COUNTRY 339 369 379 339 369 379 339 369 379 --------------------- USA . 378 306 410 85.3 86.2 93.8 111 91 387 England . 23 26 6 5.2 7.3 1.4 222 46 667 France 8 4 4 1.8 1.1 0.9 50 25 100 Germany 5 8 4 1.1 2.3 0.9 20 0 0 Other 29 11 13 6.6 3.1 3.0 34 45 15 Total 443 355 437 100.0 100.0 100.0 llO 83 374 Sampling and Short-Period Usage in the Purdue Library I 215 TABLE 4 LmRARY's HoLDINGs AND RELATIVE UsAGE BY LANGUAGE OF TITLE NUMBER s IN DDC 330- 340- 370- LANGUAGE 339 369 379 ------ English 417 338 427 French 9 4 5 German 4 10 4 Other 13 3 1 Total 443 355 437 the library three times as much as un- dergraduates and four times as much as faculty. This is in close agreement with the findings of Snyder.3 Where Were the Titles? Every librari- an would like to know where his books are: how many are checked out, how many are on shelves, etc. We carried out an inventory of titles in sample S for Education ( DDC 370-379) on July 22- 23 and found the following (Table 6). In Table 6 .. not traceable" means not accounted for by any of the previous categories. It was thought that some of these might be back4 on shelves after the summer session and accordingly two more rounds were made . on August 14 and 17 respectively. During these rounds twenty-two out of eighty-one "'not trace- able" titles were found, and the remain- ing were considered either missing or misshelved. The estimate of the avail- ability on shelves ( 65 per cent) is quite close to that of Trueswell. REGRESSION MoDELS AND STORAGE FUNCTIONS Originally, the plan was to quantify the four factors ( 1 ) language, ( 2) coun- try, ( 3) year of publication, and ( 4) year of accession. Unfortunately, this could not be done mainly because ( 1 ) 8 Helen I. Snyder, "Toward an Optimal Library System for Pennsylvania State University." Paper presented at American Society for Engineering Edu- cation annual meeting, June 1965. 4 Could be due to in-library use, in transit, records under process, etc. during July 22-23. PER CENT S IN DDC PER CENT R IN DDC 330- 340- 370- 330- 340- 370- 339 369 379 339 369 379 ------------------ 94.1 95.2 97.7 116 87 382 2.0 1.1 1.2 44 0 60 0.9 2.8 0.9 25 0 0 3.0 0.9 0.2 0 0 0 100.0 100.0 100.0 110 83 374 94 per cent or more titles were in Eng- lish, ( 2) 85 per cent or more titles were published in the USA, and consequently ( 3) there were very few, if any, titles in the samples belonging to some of the thirty classifications based on the four factors considered above (especially with "'Not English"). The best that could be done was to fit a regression model for English-language titles. The follow- ing model for the English titles in DDC 330-339 was obtained: R = 181 + 1619xl - 112x2 - 62x3 107x4 - 81x5 - 134x6 - 1538xlx3 TABLE 5 UsER Class Number Undergraduate 2,444 Graduate 3,497 Faculty 1,023 Other 3,821 Total 10,785 TABLE 6 WHERE? On shelves . Checked out . . . . On reserve or reference . On indefinite loan . Known missing in 1960 Not Traceable Total . NUMBER OF TITLES USED Per 100 Total Persons ------ 216 8.8 875 25.0 66 6.5 368 9.6 1,525 14.1 SAMPLES Number Per Cent 286 65.4 20 4.6 26 6.0 20 4.6 4 0.9 81 18.5 437 100.0 216 I College & Research Libraries • May~ 1966 TABLE 7 LmRARY' s HoLDINGS AND USAGE ( H, I, R) BY LANGUAGE, CoUNTRY, YEAR OF PUBLICATION AND YEAR oF AccEssiON NUMB ER IN SAMPLE PER CENT YEAR OF YEAR OF DDC LANGUAGE COUNTRY PUBLICATION ACCESSION s I H s R ------ USA 1954-64 Any 110 45 154 24.8 181 1944-53 Any 90 40 67 20.3 119 1924-43 Any 103 25 51 23.3 74 Up to 1923 After 1943 8 7 1 1.8 100 l English Up to 1923 Up to 1943 64 17 13 14.4 47 330-339 England Mter 1923 Any 17 25 22 3.8 276 Up to 1923 Any 6 2 2 1.4 67 Other Any Any 19 3 8 4.3 58 Not English Any Any Any 26 4 1 5.9 19 1954-64 Any 111 61 107 31.3 151 1944-53 Any 70 33 27 19.7 86 rSA 1924-43 Any 88 28 9 24.8 42 Up to 1923 Mter 1943 2 4 1 0.6 250 rnglish Up to 1923 Up to 1943 33 2 5 9.3 21 England Any Any 26 6 6 7.2 46 340-369 Other 1954-64 Any 3 2' 4 0.9 200 Up to 1953 Any 5 0 0 1.4 0 Not English Any Any Any 17 0 0 4.8 0 1954-64 Any 157 363 625 35.9 629 1944-53 Any 75 94 180 17.2 365 rSA 1924-43 Any 95 86 106 21.7 202 rnglish Up to 1923 Mter 1943 20 4 12 4.6 80 Up to 1923 Up to 1943 62 31 87 14.2 190 370-379 England Any Any 6 9 31 1.4 667 Other Any Any 12 0 3 2.8 25 Not English Any Any Any 10 0 3 2.2 30 153lxlx4 - 1669xlx5 - 159lxlx6 - 1x2xa + 238x2x4 + 12x2x5 + 6Sx2x6, It must be pointed out that since the number of titles in samples S, I, and H is rather small for quite a few of the fifteen cells used in the above regression model, the estimates of the correspond- ing regression coefficients are subject to large sampling errors. The magnitude of these sampling errors, of course, can be reduced by taking larger samples. where R = relative use; variables x1, x2 quantify country and xa, X4, X5, X6 quan- tify age as follows. Country USA . England France, Germany and Other Year of Year of Publication Accession 1954-1964 Any 1944-1953 Any 1924-1943 Any Up to 1923 Mter 1943 Up to 1923 Up to 1943 xa 0 1 0 0 0 x4 0 0 1 0 0 0 0 1 0 0 1 x5 ~ 0 0 0 0 0 0 1 0 0 1 Similar regressions have been worked out for DDC 340-369 and DDC 370-379. The library's holdings and relative usage by language, country, year of pub- lication, and year of accession are pre- sented in Table 7. The classification based on the above four factors is as detailed as possible subject to the small samples S, I, and H. This table is both interesting and useful. Relative usages in this table are the estimates of the re- gression coefficients. Based on Table 7 (or the multiple linear regression) the functions given in Sampling and Short-Period Usage in the Purdue Library I 217 Table 8 have been constructed to identi- fy Economics (DDC 330-339) mono- graph titles for possible storage. Similar functions have been constructed for DDC 340-369 and DDC 370-379. ·It is encouraging to note that the results of storage functions for DDC 340-379 are quite similar to those in Table 8 for DDC 330-339. These decision rules are relatively simple to apply and easy to communicate to the patrons. However, the evaluation of these functions has to be made by the administrators of the library with respect to its over-all goals. USAGE OF LmRARY F AGILITIES The Johns Hopkins U niversity5 has car- ried out a survey of library usage to con- struct a picture of the activities which G The Johns Hopkins University, "Progress Report on an Operations Research and Systems Engineering Study of a University Library," April 1963. make up a library day. A questionnaire survey was also conducted in a part of the Purdue general library (not restricted to social sciences only) to study the patrons' purposes of visit and prefer- ences, etc. This survey was in operation for a few hours on each day during July 21-August 7, 1964 and 212 com- pleted questionnaires were received. Table 9 summarizes the replies regard- ing the purpose of visit to the library. In ( ii) a patron has been counted more than once if he reported more than one purpose. "Use of own material," "use of library material," and "check out for home use" were mentioned by 60 per cent, 54 per cent, and 20 per cent of the persons respectively. About 8 per cent reported "other" (or personal business). According to the Johns Hopkins study about 10 per cent of the patrons con- TABLE 8 STORAGE FUNcTIONs FOR EcoNOMICS ( DDC 330-339) MoNOGRAPH TITLES Storage Function 1. Use as a function of publica- tion date ( assuming mono- tonicity) 2. Use as a function of acces- sion date ( assUining mono- tonicity) 3. Use as a function of publi- cation date excluding post 1943 accessions 4. Use as a function of publi- cation date, accession date, language and country (multi- ple linear regression) Per Cent Titles Stored Upper Cutting Point (i) 10 (ii) 20 (iii) 30 ~·) 10 ~) 20 (iii) 30 (i) 10 ( ii) 20 (iii) 30 (i) 6 ( ii) 20 (iii) 26 1915 1927 1935 1921 1934 1940 1917 1929 1936 All not in English All in ( i) ; English titles pub- lished in USA before 1924 and accessioned before 1944 All in ( ii); English titles pub- lished in England before 1924; English titles published in other countries Per Cent Use Generated by This Group 4.3 8.6 15.8 2.1 9.4 16.2 4.1 8.2 15.6 1.0 7.2 10.2 218 I College & Research Libraries • May, 1966 TABLE 9 PURPosE OF VISIT TO THE LmRARY PERSONS PART PURPOSE Number Per Cent (i) Use of own material only 0 0 69 33 Use of library rna- terial only 44 21 Check-out for home use only 7 3 Other only 0 17 8 Use of library and own material 0 40 19 Use of library rna- terial and check- out . . . . -17 8 Use of own material and check-out 5 2 Use of library and own material and check-out 13 6 Total . . . . 212 100 (ii) Use of own material 127 60 Use of library rna- terial 0 114 54 Check-out for home use 42 20 Other 17 8 ducted personal business, and about 50 per cent used their own material. When asked '1f checking out items now, did you intend when you came, to borrow them or did you get interested in them as a result of browsing?", "came to borrow," "result of browsing," and ''both" were mentioned by 59 per cent, 34 per cent, and 7 per cent respectively of those who replied to this question. About 46 per cent of library patrons "preferred" and 29 per cent "did not pre- fer" to use library material in the library rather than checking it out for home use. The reasons for those preferring the library were: better study atmosphere ( 46 per cent), to avoid mislaying of ma- terial ( 17 per cent), easier to refer to other sources ( 12 per cent), save the trouble of carrying it home ( 8 per cent), etc. Similarly, the reasons for those not TABLE 10 PERSONS REASON Number Per Cent Own interest 0 0 Course assignments 0 Needed for t erm paper Study for course exam Other 0 Total0 0 69 52 51 28 16 152 45 34 34 18 11 100 • The actual sum of "Persons" exceeds total be- cause of multiple reasons . preferring the library were: more com- fortable at home ( 40 per cent), need for longer period ( 29 per cent), can use at leisure ( 16 per cent) , use in conjunction with own material or typewriter ( 7 per cent), etc. The library patrons who used library material during "this visit," were asked the reason for use of the library material in the library. The replies were as fol- lows (Table 10). On an average a library patron spent 2.2 hours in the library and used three titles during one visit. Thus, he spent approximately 0.7 hours per title used in the library. This may be compared with the usage of 7.7 hours per title at home obtained by a survey of material used at home during the same period. CoNCLUSION This investigation was mainly explora- tory and will be used as a guide for fur- ther and more complete studies of usage in the Purdue library. As mentioned earlier, patterns of use in fall and spring may be quite different from those in summer. It is therefore advisable to ana- lyze data on use throughout the year. Also, it is necessary to take a larger sample to reduce the magnitude of sam- pling error. • •