Quality & Quantity 35: 277–289, 2001.
© 2001 Kluwer Academic Publishers. Printed in the Netherlands.
The Inﬂuence of Graphic Techniques in the
Evaluation of the Effectiveness of Treatment in
, M. L. DOMÍNGUEZ and T. JURADO
Department of Experimental Psychology, University of Seville, Spain
Abstract. The objective of this research was to see if the evaluation of the effectiveness of a given
treatment through visual inspection depended on the type of graphs used. Three subjects were studied
using 3000 graphs (1000 line, 1000 bar and 1000 box-plots), each one representing the results of
an A–B design. The results indicate that the subjects committed more errors when the data were
presented by line or bar graphs instead of box-plots.
Key words: graphical perception, methodology, time-series, visual analysis
For a long time graphic techniques have been the principal form for evaluating
the effectiveness of treatment in the area of behavioral modiﬁcation (Parsonson
and Baer, 1986). However, various authors (Matyas and Greenwood, 1990) have
maintained the inconsistencies that exist between statistical inferences and those
derived from visual inspection. They have found that the inconsistencies depend
on the statistical characteristics that exist in the data (Jones et al., 1978; Knapp,
1983; etc.), and on the subjects’ training (Wampold and Furlong, 1981).
Nevertheless, research has shed little light on how individuals visually analyze
time-series data (Furlong and Wampold, 1982). A possible explanation is that the
subjects use distinct or different types of strategies: some of them only take changes
between phases into account, while others take the data variance and do not just
base their judgement on the changes between phases (Wampold and Furlong,
1981). Undoubtedly, the ideal visual analyst performance must be like an statistical
decision test. That is to say, the analyst who believes that the evaluation of the
effectiveness of treatment consists of determining whether or not the difference in
the means between the distinct phases is sufﬁciently great. To determine this, the
visual analyst must compare this difference with the degree of existing variance in
the data. If the variance is very ample then the difference between the means will
Corresponding author: Department of Experimental Psychology, University of Seville, Sevilla,
Spain 41005, phone: 32 54 55 7683; fax: 32 54 55 1784; e-mail: firstname.lastname@example.org