Analyzing Survey Responses
You may analyze a specific survey question on-screen by selecting the survey and choosing the Analyze Survey Responses option. From there, you may choose a specific question and view or download the response data. This will include data from participants who are still in the process of completing the survey, but have not yet completed it. The Download Survey Responses option, which provides the entire set of survey responses for all participants, includes only data from participants who have completed the survey. This is done to ensure easier data analysis.
In rare cases, you may notice that a few responses for certain questions appear (or disappear from) the Analyze Survey Responses while never showing up in Download Survey Responses. In this situation, a participant has started an online survey but never fully completed it or withdrew from it (they more than likely just closed their web browser). The system goes through all “orphaned” responses in this situation and clears them out once they are more than 2 days old.
It is presumed that you will want to analyze the survey data across all questions. In this case, you should choose the Download Survey Responses option.
To successfully analyze the data, you will need to download 2 sets of data. The first is the question key, which lists a unique numeric identifier for each question, along with the question text and abbreviated question name. It also includes the section number that each question was in. Note the section number listed is merely a unique identifier for each section, and has no correspondence to the order in which sections were presented. Also note that section mean and sum values, if calculated, will normally be at the end of the list of questions. This data is in CSV format and is available by clicking the Download Question Key link on the Download Survey Responses Page.
After you download the question key, you can download the survey data. The data may be separated into parts if there is an excess of data to download. The system will adjust the size of each part based on how many questions are in the survey. A survey with a large number of questions will be broken into more parts, with fewer rows of data per part. The default option is to download the entire set of survey data, meaning all parts at once. If you receive an error when doing so, then download each part one at a time. There may be too much data in the combined download for the system to process.
You may also choose to limit the download based on date range. This is based on when the participant completed the survey. If you choose a date range, this merely filters each existing part of the download. This means that some parts may be empty if no participants whose responses would normally be included in that part completed the survey during the specified date range.
There are three options for downloading and viewing the data (both the question key and responses):
- Download as CSV File. This is the most popular option. It allows you to save the data on your computer and load it in your preferred data analysis tool.
- View On-Screen (CSV Format).This option lets you view the data on-screen instead of downloading it. The data is still displayed in CSV format, and is a good way to view the format of the data, as well as ensure all the data you expect is there. Often with larger datasets, downloading and viewing the data with some tools like Excel may make it appear as if not all the data is there, when in reality it’s a limitation of the tool and not a problem with the data download. The on-screen option is also useful if data was collected in a language that uses Unicode character sets, like Hebrew.
- View On-Screen (Table Format). This option provides an easy way to view the data on-screen (to visually analyze the data) but is not very useful for purposes of loading the data into an analysis tool.
The data is presented as one row per respondent, with each of their responses in a different column. The first row includes the column headings, and the column heading maps to the abbreviated question name. Since there is no feature to specify an abbreviated question name for the computed section sum/mean item, the system automatically assigns a unique name to those columns. The naming convention is SECXXX where XXX maps to the system’s internal section identifier for that section (section_id). You can use the question key to determine which question maps to which column. The order of columns (questions) in the data download is not necessarily the same as the order of questions in the survey. These columns are ordered based on how the data is stored internally in the system. The data also includes the exact time the participant started and completed the online survey. You can use any common data analysis tool to compute how long it took the participant to complete the survey, using this information. There is also a column listing how many minutes it took the participant to complete the survey, rounded to the nearest minute.
There will be an option to include only the numeric response portion of questions that are multiple-choice with numeric choices. This is useful if the numeric question was set up with associated text, but that text should not be included in the analysis file. One example would be where the response selected was “5 Strongly Agree” but only the number 5 (and not the associated text) is useful during analysis. If you choose this option and download data and find that the associated text is included with the numeric response, then have a closer look at the survey. More than likely the question was not actually set up as a numeric question. The easiest way to deal with this downloaded data is to use a data analysis tool to do a search and replace to change the response in your downloaded data into a numeric format. Your IT department can provide assistance with such tools.
If your survey has a lot of questions (and thus columns in the output file), you may have trouble loading the data with some spreadsheet programs. Those of which are not equipped to handle such large datasets. Your best option is to use a full-featured statistical analysis package, like SPSS or SAS. Those packages handle large CSV-format data imports with ease. Your IT department can help you to select the best tools. In particular, Excel has trouble dealing with datasets with more than 250 questions (columns).