Sometimes optical scan sheets can be used a little differently to avoid the difficulties they present. You are probably familiar with such forms from taking multiple-choice tests, and survey respondents are sometimes asked to record their responses directly on such sheets. Either standard or specially prepared sheets can be provided with instructions on their use. To answer questions presented with the several answer categories, respondents can be asked to blacken the spaces next to the answers they choose. If such sheets are properly laid out, the optical scanner can read and enter the answers directly. This method may be even more feasible in recording experimental observations or in compiling data in a content analysis.
Connecting with a Data-Analysis Program
Different computer programs structure data sets in different ways. In most cases, you’ll probably sue your data analysis program for data entry. SPSS, for example, will present you with a blank matrix of rows and columns. You can assign variable names to the columns and enter data for each case (such as the responses of each survey respondent) on a separate line. Once you’re finished, your data will be ready for analysis.
As an alternative, you can often create your data set using some other means (such as a spreadsheet or a word processor) and then import the data into the analysis program. In the case of SPSS, for example, a text file with data items separated by tabs (such as datafile.dat) can be imported and then saved in the SPSS format (Such as datafile.sav). Subsequently, you can load the data file as though it had been initially created through SPSS. Most other data-analysis programs have similar options.
Whichever data-processing method you have used, you will now have a set of machine-readable data that supposedly represent the information collected in your study. The next important step is the elimination of errors: “cleaning” the data errors may result from incorrect coding, incorrect reading of written codes, incorrect sensing of blackened marks, and so forth. Data cleaning is the process of detecting and correcting coding errors. Two types of cleaning should be done: possible-code cleaning and contingency cleaning.