## Using confidence intervals in within-subject designs (1994)

Venue: | Psychonomic Bulletin & Review |

Citations: | 185 - 21 self |

### BibTeX

@ARTICLE{Loftus94usingconfidence,

author = {Mh To G. R. Loftus and Canadian Nserc and Grant Ogp and Steve Edgell and Rich Gonzalez and David Lane and Jeff Miller and Rich Schweickert and Saul Sternberg and Geoffrey R. Loftus and Michael E. J. Masson},

title = {Using confidence intervals in within-subject designs},

journal = {Psychonomic Bulletin & Review},

year = {1994},

volume = {1},

pages = {476--490}

}

### Years of Citing Articles

### OpenURL

### Abstract

Wolford, and two anonymous reviewers for very useful comments on earlier drafts of the manuscript. Correspondence may be addressed to

### Citations

1106 |
The Visual Display of Quantitative Information
- Tufte
- 1983
(Show Context)
Citation Context ...growing that exploratory, primarily graphical, techniques are at least as useful as confirmatory techniques in the endeavor to maximally understand and use the information inherent in a data set (see =-=Tufte, 1983-=-, 1990, for superb examples of graphical techniques, and Wainer & Thissen, 1993, for an up-to-date review of them). A landmark event in this shifting emphasis was publication (and dissemination of pre... |

882 |
Exploratory Data Analysis
- Tukey
- 1977
(Show Context)
Citation Context ...and, on the basis of the pattern of sample means along with some appropriate error variance, decides either to reject or to not reject the null hypothesis. In this article, we echo suggestions (e.g., =-=Tukey, 1974-=-, 1977; Wainer & Thissen, 1993) that graphical procedures—particularly construction of confidence intervals—can be carried out as a supplement to, or even as a replacement for, standard hypothesis-tes... |

652 | Envisioning Information - Tufte - 1990 |

549 | Statistical Methods for Research Workers - Fisher |

431 |
Bayesian Inference in Statistical Analysis
- Box, Tiao
- 1992
(Show Context)
Citation Context ...at are widely used in behavioral sciences today, is based on ideas developed by Bayes (1763; see Berger & Berry, 1988, and Winkler, 1993, for clear introductions to Bayesian statistical analysis; see =-=Box & Tiao, 1973-=-, and Lewis, 1993, for extensive treatments of the Bayesian approach to analysis of variance). In the Bayesian approach, the goal is to estimate the probability that a hypothesis is true and/or to det... |

380 | An Essay towards solving a Problem in the Doctrine of Chances - Bayes - 1763 |

312 |
The Design of Experiments
- Fisher
- 1935
(Show Context)
Citation Context ...ientific basis for precise knowledge of the alternative hypothesis. In Fisher’s view, power could generally not be computed, although he recognized the importance of sensitivity of statistical tests (=-=Fisher, 1947-=-). Second, Neyman and Pearson provided a prescription for behavior—that is, for a decision about whether to reject a hypothesis. Fisher, on the other hand, emphasized the use of significance testing t... |

265 |
Statistical Principles in Experimental Design
- Winer
- 1962
(Show Context)
Citation Context ... of two basic results can occur in this design: either the three error terms are all approximately equal or they differ substantially from one another. As discussed in any standard design text (e.g., =-=Winer, 1971-=-), when the error terms are all roughly equal, they can be pooled by dividing the sum of the three sums of squares by the sum of the three degrees of freedom (which amounts to treating the design as i... |

171 | On the Problem of the Most Efficient Tests of Statistical Hypotheses - Neyman, Pearson - 1933 |

155 | High-speed scanning in human memory - Sternberg - 1966 |

99 |
On methods in the analysis of profile data
- W, Geisser
- 1959
(Show Context)
Citation Context ..., 1954). Corrections for this problem have been developed in which the degrees of freedom used to test the obtained F ratio are adjusted according to the seriousness of the departure from sphericity (=-=Greenhouse & Geisser, 1959-=-; Huynh & Feldt, 1976). Second, violation of the sphericity assumption compromises the use of the omnibus error term (and its associated degrees of freedom) when testing planned or other types of cont... |

79 | On the use and interpretation of certain test criteria for purposes of statistical inference - Neyman, Pearson - 1928 |

68 | The logic of inductive inference - Fisher - 1935 |

59 |
Some theorems on quadratic forms applied in the study of analysis of variance problems, II. Effects of inequality of variance and of correlation between errors in the two-way classification
- Box
- 1954
(Show Context)
Citation Context ...is violated (and it is arguable that this typically is the case; see, e.g., O’Brien & Kaiser, 1985), two problems arise. First, the F ratio for the test of the conditions effect tends to be inflated (=-=Box, 1954-=-). Corrections for this problem have been developed in which the degrees of freedom used to test the obtained F ratio are adjusted according to the seriousness of the departure from sphericity (Greenh... |

58 |
Things I have learned (so far
- Cohen
- 1998
(Show Context)
Citation Context ... intervals, which directly addresses this question, can profitably supplement (or even replace) the more common hypothesis-testing procedures (Loftus, 1991, 1993a, 1993b, 1993c; see also Bakan, 1966; =-=Cohen, 1990-=-). These authors offer many reasons in support of this assertion. Two of the main ones are as follows: First, hypothesis testing is primarily designed to obliquely address a restricted, convoluted, an... |

55 |
Conditions under which mean square ratios in repeated measurements designs have exact F-distributions
- H, Feldt
- 1970
(Show Context)
Citation Context ... universally recommended. For example, Hertzog and Rovine (1985) recommend estimating violations of sphericity using the measure ε as an aid in deciding whether to use MANOVA in place of ANOVA (e.g., =-=Huynh & Feldt, 1970-=-). Huynh and Feldt point out that such violations do not substantially influence the Type I error rate associated with univariate F tests unless ε is less than about 0.75. For values of ε between 0.90... |

50 |
Estimation of the Box correction for degrees of freedom from sample data in randomized block and split-plot designs
- Huynh, Feldt
- 1976
(Show Context)
Citation Context ...s problem have been developed in which the degrees of freedom used to test the obtained F ratio are adjusted according to the seriousness of the departure from sphericity (Greenhouse & Geisser, 1959; =-=Huynh & Feldt, 1976-=-). Second, violation of the sphericity assumption compromises the use of the omnibus error term (and its associated degrees of freedom) when testing planned or other types of contrasts. The omnibus er... |

43 | Statistical methods and scientific induction - Fisher - 1955 |

35 | MANOVA method for analyzing repeated measures designs: An extensive primer - O'Brien, Kaiser - 1985 |

34 |
The test of significance in psychological research
- Bakan
- 1966
(Show Context)
Citation Context ...of confidence intervals, which directly addresses this question, can profitably supplement (or even replace) the more common hypothesis-testing procedures (Loftus, 1991, 1993a, 1993b, 1993c; see also =-=Bakan, 1966-=-; Cohen, 1990). These authors offer many reasons in support of this assertion. Two of the main ones are as follows: First, hypothesis testing is primarily designed to obliquely address a restricted, c... |

29 |
The empire of chance
- Gigerenzer, Switjink, et al.
- 1989
(Show Context)
Citation Context ...s techniques in the behavioral sciences have evolved as something of an expedient compromise betweensa number of ideologically conflicting approaches to drawing conclusions from statistical data (see =-=Gigerenzer et al., 1989-=-, for a thorough discussion of this assertion). Bayesian Techniques One of these approaches, which turned out not to have a strong influence on the techniques that are widely used in behavioral scienc... |

26 |
Statistical analysis and the illusion of objectivity
- Berger, Berry
- 1988
(Show Context)
Citation Context ...echniques One of these approaches, which turned out not to have a strong influence on the techniques that are widely used in behavioral sciences today, is based on ideas developed by Bayes (1763; see =-=Berger & Berry, 1988-=-, and Winkler, 1993, for clear introductions to Bayesian statistical analysis; see Box & Tiao, 1973, and Lewis, 1993, for extensive treatments of the Bayesian approach to analysis of variance). In the... |

23 |
The Fisher, Neyman-Pearson theories of testing hypotheses: One theory or two
- Lehmann
- 1993
(Show Context)
Citation Context ...; see Gigerenzer et al., 1989, p. 103). The issues raised in the Fisher versus Neyman–Pearson debate have not been settled, and are still discussed in the statistical literature (e.g., Camilli, 1990; =-=Lehmann, 1993-=-). Nevertheless, there has been what Gigerenzer et al. (1989) referred to as a “silent solution” within the behavioral sciences. This solution has evolved from statistical textbooks written for behavi... |

19 |
Design of experiments: A realistic approach
- L, McLean
- 1974
(Show Context)
Citation Context ...s in withinsubject designs. Our reasoning is an extension of that provided by a small number of introductory statistics textbooks, generally around page 400 (e.g., Loftus & Loftus, 1988, pp. 411-429; =-=Anderson & McLean, 1974-=-, pp. 407-412). It goes as follows: A standard confidence interval in a between-subject design has two useful properties. First, the confidence interval’s size is determined by the same quantity that ... |

18 |
Inductive behavior" as a basic concept of philosophy of science
- NEYMAN
(Show Context)
Citation Context ...e of belief in the hypothesis. Competing Hypotheses In contrast to Fisher’s emphasis on inductive reasoning, a third approach to statistical inference was developed by Neyman and Pearson (1928, 1933; =-=Neyman, 1957-=-). They were primarily concerned with inductive behavior. For Neyman and Pearson, the purpose of statistical theory was to provide a rule specifying the circumstances under which one should reject or ... |

14 |
On the tyranny of hypothesis testing in the social sciences
- Loftus
- 1991
(Show Context)
Citation Context ...ere, one of us has argued that construction of confidence intervals, which directly addresses this question, can profitably supplement (or even replace) the more common hypothesis-testing procedures (=-=Loftus, 1991-=-, 1993a, 1993b, 1993c; see also Bakan, 1966; Cohen, 1990). These authors offer many reasons in support of this assertion. Two of the main ones are as follows: First, hypothesis testing is primarily de... |

14 |
Graphical data analysis
- Wainer, Thissen
- 1981
(Show Context)
Citation Context ...f the pattern of sample means along with some appropriate error variance, decides either to reject or to not reject the null hypothesis. In this article, we echo suggestions (e.g., Tukey, 1974, 1977; =-=Wainer & Thissen, 1993-=-) that graphical procedures—particularly construction of confidence intervals—can be carried out as a supplement to, or even as a replacement for, standard hypothesis-testing procedures. Before doing ... |

6 | Statistics for the Social Sciences (2nd ed - Hays - 1973 |

5 |
A priori tests in repeated measures designs: Effects of nonsphericity
- Boik
- 1981
(Show Context)
Citation Context ...f contrasts, as is often done after MANOVA is applied, the sphericity assumption is no longer in effect. Thus, a viable solution is to use the appropriate specific error term for each contrast (e.g., =-=Boik, 1981-=-) and avoid the sphericity assumption altogether.s484 LOFTUS AND MASSON The problems that result from violation of sphericity have implications for the implementation of confidence intervals as graphi... |

4 |
An apology for ecumenism in statistics
- Box
- 1983
(Show Context)
Citation Context ...judgment is called for in other solutions to the debate between the Fisher and the Neyman–Pearson schools of thought, as in the suggestion to apply different approaches to the same set of data (e.g., =-=Box, 1986-=-). Graphical Procedures Traditionally, as we have suggested, the primary dataanalysis emphasis in the social sciences has been on confirmation: the investigator considers a small number of hypotheses ... |

4 |
Essence of statistics (2nd
- Loftus, Loftus
- 1988
(Show Context)
Citation Context ...re for computing confidence intervals in withinsubject designs. Our reasoning is an extension of that provided by a small number of introductory statistics textbooks, generally around page 400 (e.g., =-=Loftus & Loftus, 1988-=-, pp. 411-429; Anderson & McLean, 1974, pp. 407-412). It goes as follows: A standard confidence interval in a between-subject design has two useful properties. First, the confidence interval’s size is... |

3 |
The test of homogeneity for 2 x 2 contingency tables: A review of and some personal opiniuons on the controversy
- Camilli
- 1990
(Show Context)
Citation Context ...quasi-Bayesian”; see Gigerenzer et al., 1989, p. 103). The issues raised in the Fisher versus Neyman–Pearson debate have not been settled, and are still discussed in the statistical literature (e.g., =-=Camilli, 1990-=-; Lehmann, 1993). Nevertheless, there has been what Gigerenzer et al. (1989) referred to as a “silent solution” within the behavioral sciences. This solution has evolved from statistical textbooks wri... |

3 | Rovine M. Repeated-measures analysis of variance in development research: selected issues. Child Dev 1985;56:787–809 - Hertzog |

3 |
Bayesian statistics: An overview
- Winkler
- 1993
(Show Context)
Citation Context ...roaches, which turned out not to have a strong influence on the techniques that are widely used in behavioral sciences today, is based on ideas developed by Bayes (1763; see Berger & Berry, 1988, and =-=Winkler, 1993-=-, for clear introductions to Bayesian statistical analysis; see Box & Tiao, 1973, and Lewis, 1993, for extensive treatments of the Bayesian approach to analysis of variance). In the Bayesian approach,... |

1 |
Bayesian methods for the analysis of variance
- Lewis
- 1993
(Show Context)
Citation Context ...behavioral sciences today, is based on ideas developed by Bayes (1763; see Berger & Berry, 1988, and Winkler, 1993, for clear introductions to Bayesian statistical analysis; see Box & Tiao, 1973, and =-=Lewis, 1993-=-, for extensive treatments of the Bayesian approach to analysis of variance). In the Bayesian approach, the goal is to estimate the probability that a hypothesis is true and/or to determine some popul... |

1 | On the overreliance of significance testing in the social sciences. Paper presented at the annual meeting of the Psychonomic Society - Loftus - 1993 |

1 | Visual data representation and hypothesis testing in the microcomputer age - Loftus - 1993 |