农业院校《试验设计与分析》课程参考资料（多变量分析方法）DUDE - A User-Friendly Crop Information System

Published online June 5,2007 DUDE:A User-Friendly Crop Information System Weikai Yan*and Nicholas A.Tinker ABSTRACT any data ical ound Timkr d Yan (200) Dist (UDE). erts data fror the con exe n e has and data for data and (v) nd co rve ith Wind f om the author queryin for achievin rm an Again e to lone-term o out the arge me th nped to olve both of thes ereby pro wo types of info data imp thro and to ano rait(varial cta200,20 Yan.2001 Yan nd ra 2002 bs s of d tabas (v)to d fo ing histo al he DUDE information system (DUDE)that can facilitate the as inker and Yan (2006) wed crop system search community. on w hich a cron info VIEWING AND EDITING A COOL DATABASE ion Lib When DUDE is opnn pane wo op )it a ata fro a c for umber of variables ( d.wn a Co and oti nd locations which ar nd a table be displayed in the Mic rface alld n the ght.One important use of the Edit mode is tha rch Centre (ECORC).Agricuu data queryOOL DC6.ECORC thier shou 6 ON Ca Corresponding author (yanw@agr.gcca) inte ity of the e and its com Published in Agron..9:1029-1033(2007). COOL DUD out opening the database using MS Access 80 WI 53711 USA 1029

DUDE: A User-Friendly Crop Information System Weikai Yan* and Nicholas A. Tinker ABSTRACT A crop information system facilitates the storage, retrieval, and utilization of historical crop performance data. Here, we report a user￾friendly crop information system entitled Data Unification and Distillation Engine (DUDE), which (i) converts data from spreadsheet formats to a relational database, (ii) allows editing of tables in the database, (iii) simplifies the construction and execution of queries from the database, (iv) formats queried data into a variety of file struc￾tures for data analyses, and (v) simplifies maintenance and corrections of data in the database. DUDE runs on any personal computer that operates with Windows 2000 or later versions of Microsoft Windows. DUDE is freely available from the authors. D ATA FROM CROP PERFORMANCE TRIALS are not only essential for achieving short-term goals of selecting and recom￾mending crop cultivars, they are also valuable in addressing long-term questions about the target environments and the crop as an integrated physiological system (Yan and Tinker, 2005; Yan et al., 2007). Two types of informatics tools are es￾sential to achieving such long term goals: tools for assembling historical crop performance trial data and tools for explor￾ing patterns within such data. The biplot analysis system (Yan et al., 2000, 2007; Yan, 2001; Yan and Rajcan, 2002; Yan and Kang, 2003; Yan and Tinker, 2005, 2006) has been de￾scribed as an effective tool for exploring historical crop per￾formance trial data. This paper describes a user-friendly crop information system (DUDE) that can facilitate the assembly and use of crop performance data. Tinker and Yan (2006) reviewed crop information systems that are currently available and described in detail an example of a relational database on which a crop information sys￾tem could be based (Fig. 1). This database structure is re￾ferred to as a Context Oriented Observation Library (COOL), which is the underlying relational database for DUDE. A COOL database has the following functionalities: (i) it ac￾commodates various types of data (numeric or nonnumeric) for a large1 number of variables (measured traits, genetic markers, treatment factors) for a large number of genotypes, from a large number of studies, years, and locations; (ii) it provides a mechanism for unifying the formats and termi￾nologies of variables, genotypes, and locations, which are typically heterogeneous among different data sources; and (iii) it simplifies the construction and execution of queries for any data subset, thereby facilitating data mining. More de￾tailed description on the COOL database structure can be found in Tinker and Yan (2006). The greatest difficulty in the use of COOL also derives from its greatest advantages, namely, its serialized data struc￾ture and its division of different data types into separate, linked tables. First, it is a difficult task to convert conventional spreadsheet data, which usually has identifiers, traits, and other measurements in parallel columns, to a COOL database, while keeping the context information properly preserved. Although this can be done manually, it involves multiple steps and is error-prone even for highly trained workers. Second, although the serialized data structure of COOL facilitates querying subsets of the data, queried data must be reformatted to parallel data before they can be used in some types of data analysis. Again, although this can be done manually, it is te￾dious and prone to error. The DUDE application was devel￾oped to solve both of these issues, thereby providing an integrated crop information system. DUDE simplifies the task of data import through the use of wizards (i) to convert data from spreadsheet format to a COOL database, (ii) to edit various tables of the database, (iii) to unify trait (variable), genotype, and location names within the database, (iv) to query subsets of data from a COOL database, (v) to save que￾ried data into required formats for data analyses, and (vi) to make corrections in the database. The DUDE application was described briefly by Tinker and Yan (2006). This paper will provide a more detailed description on an updated version of DUDE, and will formally introduce its availability to the re￾search community. VIEWING AND EDITING A COOL DATABASE When DUDE is executed, an opening panel offers two op￾tions: converting data from spreadsheets to a COOL database (Populate) or extracting data from a COOL database (Query). When the Populate button is clicked, an open file dialog will appear, asking for a Microsoft Access database with a COOL data structure, to which DUDE will be attached. When a valid COOL database is selected, an interface similar to Fig. 2 will be displayed, which provides options for viewing and editing each of the database tables and for populating the database. Each of the eight tables in the COOL database (Fig. 1) can be viewed by clicking the appropriate button on the top-left of the window (Fig. 2). If the Edit button is clicked and a table is selected, that table will be displayed in the upper left area of Fig. 2 and each field of the current record (row) in that table will be displayed in area on the right. This interface allows the table to be modified from either the upper-left area or the area on the right. One important use of the Edit mode is that heterogeneous trait, variety, and location names can be unified to facilitate data query and analysis (detailed in the data uni￾fication section below). A COOL database can also be directly edited through Microsoft Access. However, the user should be careful in deleting/adding fields of a table, as this may affect the integrity of the database and its compatibility with DUDE. It is possible to use DUDE to manage a COOL database with￾out opening the database using MS Access. 1 The current implementation of DUDE uses a COOL database in Microsoft Access file format. The theoretical limit on number of records in a MS Access table is over 2 billion. However, the size of a database is limited to 2 GB, so it is not possible to specify the maximum number of records in a given table. Eastern Cereal and Oilseed Research Centre (ECORC), Agriculture and Agri-Food Canada (AAFC), K.W. Neatby Building, 960 Carling Ave., Ottawa, ON, Canada, K1A 0C6. ECORC Contribution No. 06- 735. Received 10 Oct. 2006. *Corresponding author (yanw@agr.gc.ca). Published in Agron. J. 99:1029–1033 (2007). Software doi:10.2134/agronj2006.0280 ª American Society of Agronomy 677 S. Segoe Rd., Madison, WI 53711 USA Abbreviations: COOL, context oriented observation library; DUDE, data unification and distillation engine. Reproduced from Agronomy Journal. Published by American Society of Agronomy. All copyrights reserved. 1029 Published online June 5, 2007

1030 AGRONOMY JOURNAL.VOL 99.JULY-AUGUST 2007 Fig.1.Sch n.Each box ed byh be linked to a sinele r d in the table at the ther end of the line CONVERTING DATA FROM SPREADSHEETS TO (.mdb.Mic soft Excel (*xls).or in A COOL DATABASE Populate button (Fig.2)is clicked,an Ithe hle contains a single t be listed in a list box o DUDE-View,Edit,and Add 回凶 adsheet to database FACTORS Rus P ON S VA FACTOR VALUES Refresh Tab Transpose data Start Select Continue > Edit Populate☐cose Fie.2.Interfase of DUDE for vi atine a cool database at the edit mode

CONVERTING DATA FROM SPREADSHEETS TO A COOL DATABASE When the Populate button (Fig. 2) is clicked, an Open File dialog will appear, asking for a file that contains data to be converted into the COOL database. The data file can be in Microsoft Access (*.mdb), Microsoft Excel (*.xls), or in comma-separated values text file (*.csv) format. An Excel data file can have one or multiple tables or sheets. If the file contains a single table, it will be displayed imme￾diately in the lower-left area (Fig. 2); otherwise the tables will be listed in a list box on the left hand of Fig. 2 and the user will Fig. 1. Schematic representation of the COOL database used by the DUDE information system. Each box shows the name of a table followed by the names of the data fields (columns) that are contained in that table. Lines that join tables show the relationships (linkages) between data types. The end of a line, indicated by V, identifies that many records in that table can be linked to a single record in the table at the other end of the line (1). (Adapted from Tinker and Yan, 2006, with permission.) Fig. 2. Interface of DUDE for viewing, editing, and populating a COOL database at the Edit mode. Reproduced from Agronomy Journal. Published by American Society of Agronomy. All copyrights reserved. 1030 AGRONOMY JOURNAL, VOL. 99, JULY–AUGUST 2007

YAN TINKER:A USER-FRIENDLY CROP INFORMATION SYSTEM 1031 se tab aralle t).and in xampl of trait na be help nam a 2DE Start Select Continue > arat that Edit Popttate Close ts Fig.4.The right portion of Figon the Populate mode ext Info list f the tbe the they are not rel that He e listed in side for ag es T h appr n in 4 is whts an ead er is found in the bo box for Separate between Context info and variates d to nter a ating all tables be ore the tabl Context Info Varates/Factors ong tab is the CULTIVAR YIELD bs table.ca PLOT KG/HL by a re REP TKW HT(cm) ng value empty HD (July the s DUDE che 长ace int that is to UNIFYING TRAIT,GENOTYPE,AND LOCATION NAMES no Cancel casy unific tionof the gen s two fields by wh ueryName:Fig5).Both e

be prompted to select a table from it. The selected table will then be displayed in the lower-left area. The format of the spreadsheet is flexible, except that differ￾ent traits or identifiers must be in parallel columns. The iden￾tifiers and traits can be anything and no identifiers and traits are essential. To avoid treating different traits as the same trait, however, it is highly advisable that a trait name should con￾tain, explicitly or implicitly, three pieces of information: prop￾erty, scale (unit), and method, as suggested by the authors of the International Crop Information System (www.icis. cgiar.org, verified 8 Mar. 2007), while preferably keeping the names as short as possible. Examples of trait names are Yield_kg/ha (for yield in kilograms per hectare) and Protein%_NIR (protein concentration in percentage mea￾sured using a near-infrared analyzer). Such information will be helpful for trait name unification and data querying. Two traits with different names can be merged by giving them the same alias, but two traits with identical names cannot be sepa￾rated. Data conversion is a process whereby columns of data in the spreadsheet (the lower-left area, Fig. 2) are populated into appropriate tables of the COOL database while preserv￾ing the relationships among them. Data conversion using DUDE is initiated using the Start button (Fig. 2). DUDE will first collect all the headers in the spreadsheet and present them in the Variates/Factors side of the “Separate between Con￾text info and Variates” dialog box (Fig. 3). This list is intended to capture all measured items, that is, various traits or genetic markers or experimental factors. All other headers should be moved to the Context Info list if they describe the experimen￾tal units, tests, studies, locations, varieties, cooperators, and so forth, or they can be ignored if they are not relevant. When the headers are appropriately separated and the Header Separator closed, items in the variables/factors list box will be populated into the T_Traits table, and items in the context information list box will be listed in combo boxes on the right side of Fig. 2 (Fig. 4) for use when populating the other tables. DUDE will then guide the user through appropri￾ate steps to populate the other tables in the database, namely, T_Studies, T_Varieties, T_Locations, T_Tests, T_ExpUnits, and T_Obs, and in that order. For each table in the database (upper left area) fields are assigned from columns in the spreadsheet (lower left area) that contain appropriate context information. Only one field is essential for each table, and this field is highlighted in yellow (Fig. 4). The essential field is Name for tables T_Studies, T_Varieties, and T_Locations. It is Yr for the table T-Tests and Block for the table T_ExpUnits. Shown in Fig. 4 is when the T_Varieties table is populated, and the field Name is highlighted as essential. If no appro￾priate header is found in the combo box for a required field, then it will be assumed that all new records share a common value for that field, and the user is allowed to enter a constant value in the corresponding combo box. The purpose of popu￾lating all tables before the table T_Obs is to set up the ap￾propriate relationships among tables. When this is done, each of the observed data is then inserted into the T_Obs table, each record being uniquely defined by a record in the experimen￾tal units table and by a record in the traits table (Fig. 1). The observations will be populated individually; empty cells or cells indicated as missing values will be ignored. After a COOL database has been populated with data from a spreadsheet, the same database can be populated with ad￾ditional data from another spreadsheet in the same file or from other files. DUDE checks and prevents the same dataset from being populated again by comparing each data point that is to be populated with the values that are already in the database. UNIFYING TRAIT, GENOTYPE, AND LOCATION NAMES Crop performance trial data can be highly heterogeneous, as different investigators and the same investigators in differ￾ent times may give the same trait or genotype or location dif￾ferent names or abbreviations. An important function of the COOL database is that it allows easy unification of the trait (variable), genotype, and location names. For example, the table for traits (T_Traits) has two fields by which a record can be identified (Name and QueryName; Fig. 5). Both fields are Fig. 3. Interface of DUDE to separate the spreadsheet headers into variables and context information. Fig. 4. The right portion of Fig. 2 on the Populate mode. Reproduced from Agronomy Journal. Published by American Society of Agronomy. All copyrights reserved. YAN & TINKER: A USER-FRIENDLY CROP INFORMATION SYSTEM 1031

1032 AGRONOMY JOURNAL.VOL 99.JULY-AUGUST 2007 elect Year 75 The e e d ed OK Cancel they are the trait.This allows all dat Fig.7.DUDE Interface for limiting data to certain years. database file.When a valid file is opened.all obse to by the inv ata KGHA. in Ei T then tthe are show a or th nt trait na are same trait an are in the hat an e unif chanis the DUDE inter lata can b to (Fig.2)or dir gh Micr re invo ed in the data that are currently c d w QUERYING DATA FROM A COOL DATABASE (F5g.7 ted ve In a Select or Correct Data Fig.6.DUDE interfac

required, but the Name field is used to designate unique values at the time data is imported. The QueryName field is used to hold names for querying, and it allows duplications. While the Name field should never be altered, the QueryNames field can be manually edited to identify two records that should be queried as if they are the same trait. This allows all data about a trait to be extracted at a single query, no matter what it was originally referred to by the investigators. For example, in Fig. 5, grain yield in kg ha21 was originally entered as KG HA, YIELD_KGHA, YIELD, YIELD(KGHA), or YIELDKGHA by different investigators. Once it is confirmed that these dif￾ferent trait names are for the same trait and are in the same units, they can all be given the same value (e.g., FYIELD_) in the QueryName field. A similar name unification mechanism is available for the tables T_Varieties and T_Locations. Data unification can be performed using the DUDE interface (Fig. 2) or directly through Microsoft Access. QUERYING DATA FROM A COOL DATABASE When DUDE is started and the button for Query is se￾lected, an Open File dialog will appear, asking for a COOL database file. When a valid file is opened, all observations in the database will be displayed in a spreadsheet-like window as shown in Fig. 6. The user can then restrict the observations that are shown by selecting (or deselecting) any of the studies, years, locations, blocks (replications), varieties, data type (numerical vs. nonnumerical), variable type (traits, markers, or treatment factors), or variables that are present in the data￾base by clicking the appropriate buttons on the left of Fig. 6. Figure 7 shows how the data can be restricted to certain years. When the Select Years button is clicked, all years that are involved in the data that are currently displayed will be listed and any of the years can be selected or deselected (Fig. 7). When the Year Selection Form is closed, data dis￾played in the spreadsheet-like window will be limited to the selected years. In a similar manner, data can be limited by studies, genotypes, locations, traits, and so forth. Fig. 5. Mechanism of DUDE for trait name unification in the COOL database. Fig. 6. DUDE interface for querying a COOL database. Fig. 7. DUDE Interface for limiting data to certain years. Reproduced from Agronomy Journal. Published by American Society of Agronomy. All copyrights reserved. 1032 AGRONOMY JOURNAL, VOL. 99, JULY–AUGUST 2007

YAN TINKER:A USER-FRIENDLY CROP INFORMATION SYSTEM 1033 Select Output Format Traits in a single column(serial data) MoIL-marker dala CORRECTING DATA dat ns (for SAS) CGenotype by trait table (mean across envs) xE tables for indivdual traits icked,DUDE al ow -For each GxE Table data integrity. SYSTEM REOUIREMENTS AND AVAILABILITY CBalanced by de er that OK Cancel nd MS W Fig.8.Options for saving queried data into various formats XP Th DUDE P ith some sample data files is freely available rom ReFeRenceS the data have bee n qu disp furthe g Mar 2007 es fo and.an20a. ion systems for cro perfor lata 201.GG although it is by the 1-111 8):MOTL-marker fo OTL ma Ms Ka .GGE biplo e ma data (each marker is tre ated as at I.Freg 5.A d M for OTL the ndLR (n press s the format rqu ed b many statistic al software 5.An for dis and (v)Ger ent tal an.W Sca451004016

SAVING QUERIED DATA INTO APPROPRIATE FORMATS FOR ANALYSIS Once the data have been queried, and an appropriate data subset is displayed, the subset must be saved to a file before it can be used in further data analysis. DUDE can save the data in serial format, exactly as it is displayed, with values for all traits in a single column as shown in Fig. 6. This format is, however, not readily accessible for conventional data anal￾yses, although it is readable by the GGEbiplot software (Yan, 2001). Alternatively, DUDE allows queried data to be saved into the following formats (Fig. 8): (i) MQTL-marker format for molecular marker data. This format conforms to require￾ments for QTL mapping by the MQTL software (Tinker and Mather, 1995) and is useful if DUDE has been used to assem￾ble marker data (each marker is treated as a trait); (ii) MQTL￾phenotype format for phenotypic traits. This format conforms to requirements for QTL mapping using the MQTL software (Tinker and Mather, 1995); (iii) Traits in parallel columns. This is the format required by many statistical software packages; (iv) Genotype-by-trait table averaged across replications and environments; and (v) Genotype-by-environment table for each trait. When this option is selected, the user has options to save all queried data, balance the subset by deleting environ￾ments with missing cells, or balance the subset by deleting genotypes with missing cells. Each data point will be the arith￾metic mean across replications. CORRECTING DATA Once a COOL database has been populated, it is difficult to pinpoint and correct an observation. Alternatively, DUDE provides an easy way to correct data in a COOL database. When the Correct Data button (Fig. 6) is clicked, DUDE allows the user to choose a data point specific to a year, location, genotype, replication, and trait (variable) and to give it a cor￾rect value. This is an edit-only function and does not involve data integrity. SYSTEM REQUIREMENTS AND AVAILABILITY The DUDE package is about 2.5 MB. It runs on any per￾sonal computer that is installed with Windows 2000 or later versions of Microsoft Windows. It should work on all versions of MS Excel and MS Access files but our experience is limited to MS Office 2000 and Office XP. The DUDE package, along with some sample data files, is freely available from the authors. REFERENCES Tinker, N.A., and D.E. Mather. 1995. Methods for QTL analysis with progeny replicated in multiple environments. J. Quant. Trait Loci 1(1). Available at http://wheat.pw.usda.gov/jag/papers95/paper195/ indexp195.html [verified 9 Mar. 2007]. Tinker, N.A., and W. Yan. 2006. Information systems for crop perfor￾mance data. Can. J. Plant Sci. 86:647–662. Yan, W. 2001. GGEbiplot—A Windows application for graphical analysis of multienvironment trial data and other types of two-way data. Agron. J. 93:1111–1118. Yan, W., L.A. Hunt, Q. Sheng, and Z. Szlavnics. 2000. Cultivar evalua￾tion and mega-environment investigation based on the GGE biplot. Crop Sci. 40:597–605. Yan, W., and M.S. Kang. 2003. GGE biplot analysis: A graphical tool for breeders, geneticists, and agronomists. CRC Press, Boca Raton, FL. Yan, W., S.J. Molnar, J. Fregeau-Reid, A. McElroy, and N.A. Tinker. 2007. Associations among oat traits and their responses to the envi￾ronment. J. Crop Improve. 20: (in press). Yan, W., and I. Rajcan. 2002. Biplot evaluation of test sites and trait relations of soybean in Ontario. Crop Sci. 42:11–20. Yan, W., and N.A. Tinker. 2005. An integrated system of biplot analysis for displaying, interpreting, and exploring genotype-by￾environment interactions. Crop Sci. 45:1004–1016. Yan, W., and N.A. Tinker. 2006. Biplot analysis of multi-environment trial data: Principles and applications. Can. J. Plant Sci. 86:623–645. Fig. 8. Options for saving queried data into various formats. Reproduced from Agronomy Journal. Published by American Society of Agronomy. All copyrights reserved. YAN & TINKER: A USER-FRIENDLY CROP INFORMATION SYSTEM 1033