80 likes | 114 Views
PA430 - Data coding. March 7/8, 2000. Codebook. Purpose Guide data coding process Guide to locate variables and interpret findings during data analysis Given the code book and data base, any researcher should be able to understand your data collection and do analysis using the data.
E N D
PA430 - Data coding March 7/8, 2000
Codebook • Purpose • Guide data coding process • Guide to locate variables and interpret findings during data analysis • Given the code book and data base, any researcher should be able to understand your data collection and do analysis using the data
Preplanning • As for all stages of research, pre-planning of data coding saves you many headaches later. • Issues • Type of analysis • Availability of computer applications • Other resources: time, expertise, computer • Record storage
Preplanning • It is advisable to lay out a draft of the codebook, data coding sheet, etc. before collecting data • Surveys • edge coding • consistency • Choice of computer application • spreadsheet vs SPSS
Codebooks • Key elements • introductory information including researcher’s name and institution, purpose of data collection, name of data base, etc. • variable names (3-8 characters) • sometimes a numerical code is used instead • full definition/description of variable or wording of question • variable attributes (response set) with codes
Data Cleaning • All coders will make mistakes • Sometimes even necessary for data obtained from other sources - always check • Plan for data cleaning should involve multiple steps/re-checks
Helpful hints • Save file every 10 minutes or so as you code • Keep more than one copy of the data (i.e., on a floppy and on a hard drive) • Save a copy of the original data. If you recode, save the file under a new name. • Save original surveys, coding sheets, etc. for a reasonable period of time
Data cleaning • Step 1 - “eyeball” the data • Step 2 - run frequency table for each variable • look for “impossible” response codes • Possible response 1,2,3, and you find a 4 • look for unexpected patterns • I.e., all cases coded the same for a variable • Step 3 - run descriptive statistics for each variable • again, looking for inconsistencies