| Title Page
Executive Summary
Chapter I: Problem Statement Chapter II: Shelf Reading and Type of Research Chapter IV: Reliability and Validity Chapter V: Summary
|
Data Analysis Data analysis consisted of charts on all observations and historical data. These were used to make comparisons of each aspect relating to quality. First, all data were collected and charted. Second, the mean or average of each set was calculated. Third, some of the data were analyzed through the use of chart functions in PowerPoint for comparison. A Z-test was performed to test if there were a significant difference between the shelving quality of books versus journal collections. Historical data and observational data from the actual population were collected. These provided accurate and reliable results. Measured was the exact level of quality or accuracy of items shelved in the book and journal collections. These results also provided information useful to management for making decisions regarding staffing issues. By using several different sources of information the possibility of bias was significantly reduced. Actual population data were collected through shelf-reading of the book and journal collections. This was done by experienced staff with an extremely accurate and reliable indicator of current quality. These data were useful for making comparisons to see if there were a significant difference between the two collections. The scope and limitations of this study was shelving quality or accuracy of the book and journal collections in the Greenblatt Library. The study did not include information from other collections in the Greenblatt Library nor did it include data from other health science libraries
Observational data were collected for the book collection in November 1997 and for the journal collection in August 1997. Both collections were shelf-read by experienced full-time staff and senior level library assistants. The table below displays the actual number of units each reader read (1 unit = 7 shelves), the number of errors found, and the totals. Table 1
The book collection was read by eight readers. The instructions given to readers were to count only errors from items found on incorrect shelves; however, readers 7 and 8 included all errors. These included transpositions on the same shelf. This could have skewed the results of this data collection section, slightly in the direction of having lower quality. However, because the results were already within the acceptable level of greater than 95% accuracy, no adjustments were made. The book collection was divided into units and assigned to library staff for shelf- reading. There were 232 units with seven shelves each. This was a total of 1,626 actual shelves of books. The total number of books is approximately 68,808. This figure was from the FY 1997 Annual Report Statistics kept by Cataloging Services. This was the most accurate figure available. The mean was calculated by subtracting the number of errors from the total number of books in the collection (68,808-373 = 68,435). This gave the total number of accurately shelved books. The number of accurately shelved books was divided by the total number of books in the book collection (68,435/68,808 = .9945). Lastly, this number was multiplied by 100 to give the percent of accuracy (.9945*100 = 99.45%) for the book collection. Chart 1 depicts the accuracy level of the book collection. Chart 1
The journal collection was read by three readers in August 1997. The reading was done by two experienced full-time staff and one senior level library assistant. The journal collection was divided into three approximately equal size sections. Errors were counted only if a journal was in the wrong section and errors which were transpositions within a title were not counted; e.g., volume 1, 2, 4, 3, etc., on the same shelf would be considered correct. It took a total of 24 hours to read all of the journal collection and only three errors were found. The total number of volumes were provided from the FY 1997 Annual Report. Table 2 below shows the data collected from the journal collection reading done in August 1997. Table 2
The journal collection calculations for finding the mean were done in exactly the same way as for the book collection. The total number of errors was subtracted from the total number of books in the collection (102,055 - 4 = 102051). This gave the total number of accurately shelved journals. Next, the number of accurately shelved journals was divided by the total number of books in the book collection (102,051/102,055 = .9999). Lastly, this number was multiplied by 100 to give the percent of accuracy (.9999*100 = 99.99 %) for the journal collection. Chart 2
The data from this first section were used to test the first hypothesis. It is stated as follows:
Hypothesis 1 is rejected because the mean of the book collection (99.45%) and the journal collection (99.99%) readings were greater than 95% accurate. Chart 3 shows the combined actual shelf reading comparison of the book collection versus the journal collection readings which supports these findings. Chart 3
These data were also used to test the second hypothesis which follows: Hypothesis 2: There is a difference in the quality of shelving of books versus journals. These particular data were chosen for the Z-test because they were highly valid, reliable data and were from a large quantity of data (actual population data). They were collected by experienced staff and the possibility of bias was smaller due to having several staff reading the collection. The data were used to calculate this test..
Problem: Is a .0054% difference in the mean from the book collection and the mean from the journal collection statistically significant?
H0: P1 P2 (There is a difference between the shelving quality of books versus journals) H1: P1 = P2(There is no a difference between the shelving quality of books versus journals) If = .05, and p = percentage of shelving quality/accuracy of the book and journal collections and p1 = .9945 (books), p2 = .9999 (journals), n1 = 68,808, and n2 = 10,2055 then P* = n1 P1 +n2 P2 /n1 +n2 or 68808*.9945 +102055*.9999/68,808+102,055 = .9977
The Z value at .05 level is 0.0000 for a two-tailed test. Because Z0 = Z /2 it is concluded that there is no significant difference between the shelving quality of books versus journals.
Direct shelf reading observations in the book and journal collections were done. They produced approximately the same results as actual shelf reading from population data. These results were 99.45% accuracy for books and 99.63% accuracy for journals as compared to section 1 results of 99.45% accuracy for books and 100% accuracy for journals. The spot checks were done by the LInC evening/weekend supervisor over a two months period. There were 25 random readings from each collection. This information was useful as a verification of reliability and validity. Tables 3 and 4 are the actual data sample results as collected. Table 3 displays the book collection data and Table 4 displays the journal collection data. Additional data collected from these tables include the mean for books per shelf and the mean for journals per shelf. Charts 4 and 5 show the level of accuracy for the book and journal collections separately. Chart 6 shows the comparison of the two collections. These charts indicate the collection accuracy is > 95% for both collections, thus supporting the first hypothesis and varifying the first section full shelf reading totals. Table 3
Table 4
Chart 4
Chart 5
Chart 6
Cumulative supervisor spot checks of the shelving quality of each part-time library assistant were collected from LInC evening/weekend supervisor. Ten books from each cart were chosen. This was done by the process of selecting every third book and then scanning the barcode into the Library Information System. Next the book title and call number were printed and the number of books per cart, cart number and date were noted on the paper. The cart was processed for shelving and the library assistant would sign out the cart for shelving. This was the usual process for shelving. The shelver did not know if their particular cart was being checked for accuracy. However, the shelvers did know the supervisor checked shelving accuracy on a random basis throughout the year. The LInC evening/weekend supervisor then waited until the cart was shelved and checked the 10 books for accuracy. Table 5 below displays the results. Table 5
Mean for books per cart = 47 The results from these tables were also used for testing the first hypothesis. Greenblatt Library’s book and journal collection quality of shelving is > 95% accuracy. The mean for book errors was .012 and the mean for journal errors was .000. Accuracy percentages were derived by dividing the number of errors in each section by 180 which was the number of items used to check each collection. In order to find the percent of accuracy each total was subtracted from one, then multiplied by 100. For the book collection this total was 98% accuracy and for the journal collection the total was 100 % accuracy. The results of this section are displayed in Charts 7, 8 and 9. Chart 7 depicts the accuracy level for shelving books, Chart 8 for journals and Chart 9 compares the two. Chart 7
Chart 8
Chart 9
There were zero complaint forms received which related to shelving issues. This, too, indicates high quality shelving in both book and journal collections. Dissatisfied users = .00 Satisfied users = 100%
Item locate requests for the past six months were counted and divided into groups. They were counted as possible shelving errors if they were found shelved incorrectly or not found at all. They were not counted in the study if they were found shelved correctly or checked out. There was a total of 21 locate requests for books from April 1997-September 1997. Of those 21 requests, 2 were checked out, 5 were found in the stacks and 1 was found in library. This left only 13 possible shelving errors in six months. Journal locate requests were handled in the same manner. There were a total of 19 locate requests and of those 19, 1 journal was found checked out and 12 were found in the stacks, leaving only 6 possible shelving errors for the same six-month period. There are a number of reasons other than misshelving errors which could cause a book to be missing (e.g., the book could be lost or hidden by a library user). However, these data will be useful for a later comparison to see if there is an increase/decrease in locates and compared with the total check outs. It can also be used as an indicator of quality from the perspective of users. This relates to the user being able to find or access the item they are looking for in the library. The number of check-outs for this same six- month period of time (11,560) divided into the total possible misshelvings gave a percentage of non-accessability for both collections (29/11,560=.0025). Then, the non-accessability total (.0025) subtracted from 1 and multiplied by 100 gave the percentage of accessability which was 99.75% for books and journals.
The total number of requests from April 1997 through September 1997, from full-time staff to the LInC supervisor regarding particular sections which appear to need shelf- reading or straightening was two for the book collection and zero for the journal section. This, too, was an indicator of shelving quality, but it is a better figure to follow over time as a benchmark for quality.
From April 1997 through September 1997, 10.5 hours of shelf-reading were recorded for the book collection and 4.6 hours were recorded for the journal collection. These figures are specifically library assistant shelf reading and do not include the complete shelf-reading done by the full-time LInC staff and senior library assistant. They are to be used as benchmark totals useful to follow over time.
Additional information and samples gathered included the number of minutes it took to shelf read a unit of books (7 shelves) and a unit of journals (6 shelves), the number of units in the book (236) and units of journals in the collection, and the price per hour paid to FY97 library assistants ($5.23). These totals were needed to calculate how much time it should take for the Library Assistants to shelf read both collections and how much it would cost. It was found that the mean for the number of minutes it took to shelf read one unit of books was 6.20. The average number of minutes to shelf read one unit of books times the number of units in the book collection gives the number of minutes it would take to shelf read the entire book collection (6.20 * 236 = 1,463.20 minutes or 1,463.20/60= 24.38 hours). The number of hours multiplied by the pay per hour for library assistants is the cost of reading the book collection (24.38 * $5.23 = $127.54). This same process was repeated for the journal collection. The average number of minutes to shelf read one unit of journals times the number of units in the journal collection gives the number of minutes it would take to shelf read the entire journal collection (3.04 * 754 = 2,292.16 minutes or 2,292.16/60 = 38.20 hours). The number of hours multiplied by the pay per hour for library Assistants is the cost of reading the journal collection (38.20 * $5.23 = $199.79). Table 6 contains 25 samples of the number of minutes it took to shelf read in the book collection and Table 7 contains 25 samples of the number of minutes it took to shelf read in the journal collection. Table 6 Each number in this data set represents the number of minutes it takes to read 1 range of books Column 1 Column 2 Column 3 Column 4 Column 5
mean = 163 (total number of minutes spent) / 25 (total number of readings) = 6.52 minutes per range Table 7 Each number in this data set represents the number of minutes it takes to read 1 range of journals Column 1 Column 2 Column 3 Column 4 Column 5
mean = 76 (total number of minutes spent) / 25 (total number of readings) = 3.04 minutes per range The alternative to the null hypothesis one was accepted because the data supports that there is a greater than 95% accuracy for the book and journal collections at the Greenblatt Library. In fact, the data support a greater than 99.72% accuracy level average for the two collections. The alternative to the null hypothesis two was accepted because the data results from the Z-test found there was no significant difference between the quality/accuracy of the book collection versus journal collection.
|
Copyright © 1998 |
Library
| Medical College of Georgia |
