| The time visitors spend engaged in an activity is a key measure of visitor behavior in museum studies. This study examined the distribution characteristics of visitor time data and studied various descriptive and comparative statistical methods for its analysis. Among the statistical methods studied was survival analysis, which had not previously been applied to visitor time data.;Two sets of time data were analyzed. One set consisted of the times of 100 visitors in an education gallery under two gallery conditions. The other set consisted of the times 100 visitors spent looking at one work of art. The looking time of accompanied and unaccompanied visitors was compared.;The time distributions in both studies were positively skewed and had considerable variability. The findings of the study suggest that a close examination of the distribution of visitor times is essential, so that the shape and spread of the distribution is clearly identified. The median is the most appropriate point estimate of central tendency of a skewed visitor time distribution; however, due to the variability of visitor time data, a single point estimate such as the median or mean does not adequately describe "typical" visitor behavior. The survival curve is useful for data description because it presents a "whole pattern image" of visitor engagement behavior over time.;When the mean times of two or more independent samples are compared, the study should be designed so that the samples are equal in size. Time distributions are likely to be nonnormal and have heterogeneous variances, and parametric tests such as the analysis of variance are not robust when sample sizes are unequal. A nonparametric, distribution-free test such as the Mantel-Cox test is an alternative with many advantages. In particular, it tests the equality of two or more cumulative survival functions, and is a whole pattern test rather than a test of a point estimate; thus, it is sensitive to group differences across the time continuum. |
