Estimates of cross-validity for stepwise regression and with predictor selection.

The effects of preselection of predictors (e.g., stepwise regression) on formula estimates of cross-validity were examined. Three actual data sets were used to generate populations of varying sample size, population validity, and number of predictors. No formula estimate provided an unbiased estimate of the population cross-validity, although some formula estimates were less biased than others. More important, having an adequate sample size (relative to number of predictors) was the issue most affecting the utility of the formula estimates. Another conclusion was that adjusted R2 provided by at least some popular software programs can provide gross overestimates of cross-validity and should not be used as such. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

A SAS macro for stepwise correlated binary regression

Changes in complete blood count (CBC), bone marrow, reticulocyte count, coagulation profile, biochemical analysis, and serum iron, transferrin, and immunoglobulin M (IgM) concentrations were measured three and 10 days after splenectomy in 12 dogs. Spleens were cultured aerobically for bacteria and submitted for histopathological evaluation in 23 dogs undergoing splenectomy. There were no consistent changes in red blood cell (RBC), white blood cell (WBC), or platelet counts; bone-marrow samples; or biochemical profiles. Serum iron, transferrin, and IgM concentrations remained normal. Eight (35%) bacterial cultures yielded growth. Five of the 23 dogs had pyrexic episodes two-to-five days after surgery. In contrast to previous reports done on healthy dogs, this study shows that dogs with splenic disease have no characteristic changes in hematological or biochemical parameters after splenectomy. Rather, the changes tended to reflect the primary disease process. Splenic vascular compromise or a decrease in processing of bacteria may have resulted in the bacterial growth. There was no direct correlation to pyrexic postoperative episodes.     

基于分步回归策略的霍脱尔气体发射率数学模型

针对冶金炉窑燃烧释热场、温度场计算机仿真过程中气体发射率的动态计算问题,采用计算机图形处理技术从霍脱尔发射率曲线图中智能提取数据,并基于分步回归策略,建立了CO2、SO2、H2O等气体的发射率数学模型.模型检验和误差分析表明,该模型回归结果误差小,满足辐射换热工程计算精度要求.     

Reduced-variance regression.

Suggests that the most useful statistical techniques for practical prediction problems, as in personnel selection, may be 2 new techniques, ridge regression and Stein-type regression. They combine the sensitivity of multiple regression with the resistance to sampling error of other techniques—notably rational (clinical) weights and weights determined by simple correlations. However, the new techniques are not recommended for theoretical modeling work, since they yield biased estimates of the true least squares weights, typically have higher expected squared errors for estimating some weights, and do not allow the use of ordinary confidence bands or significance tests. (24 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Tests of serial scanning in item recognition.

Conducted 2 experiments to examine S. Sternberg's (see record 1966-10810-001) theory that humans scan a memorized set of items serially and exhaustively to determine whether it contains a given probe item. Eight undergraduates memorized sets of 2, 4, 6, 8, 12, and 16 items and were presented with a series of 96 probes for each set. Although mean RTs increased with set size in a manner largely consistent with serial scanning, analysis of RT distributions showed that increasing set size increased the positive skew of the distribution. The slope of the function relating mean RT to set size was influenced by repetition of negative probes in the series of probes and was higher when a probe was a 2-syllable non-noun than when it was a 1-syllable noun. The latter effect was independent of whether the sets consisted wholly of nouns or of non-nouns or whether they were mixtures of the two. Findings support the serial-scanning model, which indicates that Ss scan on only a proportion of trials. (French abstract) (32 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Tests for success in diver training.

46 20–30 yr old male US Navy personnel were tested with measures from 5 physical fitness batteries prior to beginning a rigorous physical and diving training program in order to develop a new model of physical performance. Results show that cardiovascular fitness, muscular endurance, and flexibility were most highly related to finishing the program. Findings highlight the need for a more inclusive conceptual framework for understanding and predicting performance in a wide range of occupational categories. (19 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Tests of dominance in sequential categorical data.

Presents procedures to test research questions related to dominance in cases such as marital and family interactions. Dominance is an important phenomenon in interactive patterns and is defined as an asymmetry in predictability of 2 members of a system. Unlike previous tests for dominance that are applicable to categorically coded data, these tests account for the dependence between the 2 unidirectional tests. (6 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Hypnotic age regression.

This review covers the types of measures which have been used to compare the test performance and behavior of the S in the waking state with that in the hypnotic state, the various conditions under which the performance is recorded, the principal established results, the main theories which account for the phenomenon, and the methodological problems involved. Suggestions for future research are made. From Psyc Abstracts 36:01:3II29Y. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Multiple monotone regression.

Multiple monotone regression is a set of models and methods that can be applied when researchers do not think that the relationships among their measured variables can be adequately described by a multiple linear regression model. The relations among weak monotonic, additive, and additive monotonic models are discussed. As an example, the models are applied to data from the present author's (see record 1982-27256-001) study of visual information processing. (17 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Tests and college admissions.

Discusses survey data indicating the different emphases placed by institutions of higher education on a range of admissions criteria including academic performance, evidence of graduation, letters of recommendation, personal essays, and interviews. It is concluded that standardized tests are not the pervasive barrier to higher education in the US that many test critics claim. (8 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Attempted suicide among adolescents: A stepwise discriminant analysis.

A sample of 40 adolescents (aged 13–17 yrs), some of whom had attempted suicide (n?=?20) and some who had not (n?=?20), were compared on a number of life history and psychological variables (affective and cognitive). Stepwise discriminant analyses produced 1 discriminant function (the lack of emotional significant other) that differentiated between the groups. This single discriminant function accounted for 77% of the variance in the data. These results are interpreted within the context of the early loss hypothesis. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Tests of the list-strength effect in recognition memory.

The list-strength effect arises when increasing the strength of some items in a list reduces memory for the remaining items. The list-strength effect was investigated under conditions of rapid visual presentation. Randomized and blocked formats were used for the mixed lists. Performance was measured with both yes–no and forced-choice recognition procedures. Overall no evidence for a list-strength effect in recognition was found except under conditions that may promote reverse rehearsal borrowing. Two experiments were conducted to determine why performance on the yes–no tests was greater than on the forced-choice tests. Repeated testing with the yes–no procedure promoted more effective encoding than the forced-choice procedure. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Tests for homogeneity of variance in factorial designs.

The variance portion of P. A. Games's (See PA, Vol 62:2685) 3-factor model of inference on independent groups is extended. Six procedures that convert tests of spread into tests of location are reviewed and explored in a Monte Carlo study of how to test variances in a factorial design. The statistic In s–2 is shown to be a slight improvement over the procedure outlined by J. E. Overall and J. A. Woodward (see record 1975-04313-001). The dependence of these 2 tests on the normality condition is illustrated. Four robust alternatives of somewhat lower power are contrasted. The jackknife test is the most powerful and is only slightly sensitive to leptokurtosis if the ns are equal. The median test described by M. B. Brown and A. B. Forsythe (1974) is acceptable, but it uses average deviations rather than variances. The Box-Scheffé test is always robust. No single test is ideal. A 2-stage process is recommended. (31 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Tests of lagged dominance in sequential dyadic interaction.

Points out several shortcomings of the estimation and testing techniques proposed by P. D. Allison and J. K. Liker (see record 1982-12033-001). An improved approach is presented in which parameters of interest are estimated under the best fitting multivariate linear model of the dyad behavior, and the test for dominance is obtained by imposing appropriate linear constraints on the vector of estimated parameters. An additional type of dominance possible in dyadic interaction—indirect dominance—is defined, and a typology of dominance structures is presented. Two numerical examples are analyzed by the proposed technique. (30 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Empirical evaluation of ridge regression.

Evaluated ridge regression by computer simulation over a variety of parameter conditions. Results show that ridge improved upon ordinary least squares only for a limited and predictable set of parametric conditions and did not provide a general antidote to the error variance accompanying multicollinearity. Ridge was effective for conditions of low R–2 (signal to noise ratio) and, for analyses based on the principle components of predictors, when principal components with small eigenvalues made only a small contribution to predicting the dependent variable. Routine use of ridge regression without prior knowledge of the predictor space is not recommended. Derivation of the orientation parameters is appended. (34 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

A reconceptualization of adaptive regression.

Contends that the concept of adaptive regression is useful because it emphasizes that regression can be adaptive, but is misleading in suggesting that adaptive regressions are a discrete and unitary set of phenomena. Recently, writers have stated that primary process and creativity are not regressive, but these ideas imply an implausibly sharp break between regressive and adaptive phenomena, and are contradicted by empirical observations. The author notes that regression may be adaptive in a variety of ways: It may be chosen and used for communication or for privately adaptive ends, or it may occur unbidden, but nevertheless, be used for communication or privately adaptive ends. It is suggested that a more meaningful question than "Is regression adaptive or maladaptive," is "How is regression adaptive, maladaptive, or both?" (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Tests enhance the transfer of learning.

Numerous learning studies have shown that if the period of time devoted to studying information (e.g., casa-house) includes at least 1 test (casa-?), performance on a final test is improved—a finding known as the testing effect. In most of these studies, however, the final test is identical to the initial test. If the final test requires a novel demonstration of learning (i.e., transfer), prior studies suggest that a greater degree of transfer reduces the size of the testing effect. The authors tested this conjecture. In 2 experiments, 4th- or 5th-grade students learned to assign regions or cities to map locations and returned 1 day later for 2 kinds of final tests. One final test required exactly the same task seen during the learning session, and the other final test consisted of novel, more challenging questions. In both experiments, testing effects were found for both kinds of final tests, and the testing effect was no smaller, and actually slightly larger, for the final test requiring transfer. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Distortions of construct validity in multiple regression analysis.

Contends that interpreting the results of multiple regression analyses in terms of the constructs reflected by variables originally entered into the analyses is hazardous because the statistics of interest produced by the analysis (e.g., regression coefficients) are computed using residualized scores. These scores reflect constructs that differ from those described by the original raw scores. Interpretive hazards are demonstrated with a simple fabricated set of data and then with actual data from a study by the author and R. W. Marx (1981) relating 118 schoolchildren's self-concept to their achievement. It is recommended that anchor variables not of direct interest in a research study be measured and correlated with residualized variables. This supplementary analysis sheds light on changes to construct validity that must be known before interpreting multiple regression analyses. (French abstract) (19 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Tests of attentional flexibility in listening to polyrhythmic patterns.

Four experiments examined attentional flexibility in listening to polyrhythmic patterns. Musically trained and untrained listeners detected changes in timing of 1 tone (the lower tone) in a 3:2 polyrhythm in which high and low tones varied in frequency separation. Experiment 1 encouraged integrative attending; all listeners performed significantly poorer in conditions with wide as opposed to narrow frequency separations. Experiment 2, which encouraged selective attending to low tones, reversed these results: Performance was poorer in the narrow frequency conditions. In neither experiment did skill interact with frequency separation. Experiments 3 and 4 extended these findings to moderate frequency separations. Over all experiments, musically trained listeners exhibited an enhanced ability to detect timing variations, but not flexibility of perceptual organization as it applies to detection of timing changes. Instead, pattern structure (e.g., frequency and time relation) decisively influenced perception for both levels of skill. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

The Power of Statistical Tests for Moderators in Meta-Analysis.

Calculation of the statistical power of statistical tests is important in planning and interpreting the results of research studies, including meta-analyses. It is particularly important in moderator analyses in meta-analysis, which are often used as sensitivity analyses to rule out moderator effects but also may have low statistical power. This article describes how to compute statistical power of both fixed- and mixed-effects moderator tests in meta-analysis that are analogous to the analysis of variance and multiple regression analysis for effect sizes. It also shows how to compute power of tests for goodness of fit associated with these models. Examples from a published meta-analysis demonstrate that power of moderator tests and goodness-of-fit tests is not always high. (PsycINFO Database Record (c) 2010 APA, all rights reserved)