Web Survey Bibliography

Title Visual Analogue Scales: Non-linear Data Categorization by Transformation with Reduced Extremes
Author Funke, F., Reips, U. -D.
Source General Online Research Conference (GOR) 2006
Year 2006
Access date 20.09.2006
Abstract Two different ways of analyzing data from visual analogue scales (VAS) have been compared to one another: linear transformation that is commonly used, and transformation with reduced extremes. The VAS that have been examined here consist of horizontal lines with verbal anchors at both ends. VAS values can be determined accurate to a pixel and fall into the category of continuous response formats. By generating data on interval scale level, they meet important requirements for the applicability of parametric procedures. In two online experiments the influence of VAS on the respondent's way of answering questions has been examined in comparison to (4-, 5-, and 8-point) categorical scales.

As expected, the two response formats had no influence on central statistical parameters. Although, on the face of it, they appeared to be equivalent, it could be shown that the commonly used linear transformation of VAS' values results in systematic differences. This means that the use of different response formats can lead to different interpretations. An alternative way of categorizing data obtained from VAS is therefore suggested, and in the first experiment it could be demonstrated that by using this alternative data from both response formats turned out to be actually equivalent.

To be able to compare the frequencies of categorical scales with VAS, segments of the VAS have to be put together to form categories. Two ways of categorizing data have been examined. The first one consists in linear transformation, where equal intervals of the VAS are put together to form one category. This leads to equally spaced categories with same size and distance and represents the transfer of the VAS' characteristics onto categorical scales. The second way is the use of transformation with reduced extremes. The segments on the VAS that are put together to form the extreme categories are smaller here than the segments that form the other categories. Transformation with reduced extremes leads to greater correspondence with the frequencies on categorical scales and, all in all, is clearly superior to linear transformation. Since on VAS the extreme categories are represented by smaller intervals than the other categories, they appear to be less important to the respondent. This means that the centers of categories and categorical scales have different distances so that they are not perceived as equidistant.

In a second experiment - to take into account the reduced distances of the extreme categories - the categorical scales were optically modified according to the transformation procedure with reduced extremes. The space between the radio buttons (i.e. categories) is no longer the same but represents the reduced extremes. The distance between the outermost button and the adjoining one is smaller than the distance between the other buttons. The aim is to create a scale whose categories are perceived as having the same distances so that the required equidistance is no longer violated.
Access/Direct link Conference homepage (abstract)
Year of publication2006
Bibliographic typeConferences, workshops, tutorials, presentations
Full text availabilityNon-existant
Print

Web Survey Bibliography - Funke, F. (36)