Another look at ANCOVA versus blocking.

Compared methods of incorporating a concomitant variable into an experimental design. A 3?×?3 conceptual framework for these methods was developed, with one dimension representing 3 methods of assignment to treatment groups and the other dimension representing 3 methods of data analysis. Monte Carlo procedures were used to investigate the relative statistical power and apparent imprecision of the 9 methods. Results show that the recommendation of most experimental design texts to consider the correlation between the dependent and concomitant variables in choosing the best technique for utilizing a concomitant variable is incorrect. Instead, the 2 factors that should be considered are whether scores on the concomitant variable are available for all Ss prior to assigning any Ss to treatment conditions and whether the relationship of the dependent and concomitant variables is linear. (32 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Issues for covariance analysis of dichotomous and ordered categorical data from randomized clinical trials and non-parametric strategies for addressing them

Analysis of covariance is an effective method for addressing two considerations for randomized clinical trials. One is reduction of variance for estimates of treatment effects and thereby the production of narrower confidence intervals and more powerful statistical tests. The other is the clarification of the magnitude of treatment effects through adjustment of corresponding estimates for any random imbalances between the treatment groups with respect to the covariables. The statistical basis of covariance analysis can be either non-parametric, with reliance only on the randomization in the study design, or parametric through a statistical model for a postulated sampling process. For non-parametric methods, there are no formal assumptions for how a response variable is related to the covariables, but strong correlation between response and covariables is necessary for variance reduction. Computations for these methods are straightforward through the application of weighted least squares to fit linear models to the differences between treatment groups for the means of the response variable and the covariables jointly with a specification that has null values for the differences that correspond to the covariables. Moreover, such analysis is similarly applicable to dichotomous indicators, ranks or integers for ordered categories, and continuous measurements. Since non-parametric covariance analysis can have many forms, the ones which are planned for a clinical trial need careful specification in its protocol. A limitation of non-parametric analysis is that it does not directly address the magnitude of treatment effects within subgroups based on the covariables or the homogeneity of such effects. For this purpose, a statistical model is needed. When the response criterion is dichotomous or has ordered categories, such a model may have a non-linear nature which determines how covariance adjustment modifies results for treatment effects. Insight concerning such modifications can be gained through their evaluation relative to non-parametric counterparts. Such evaluation usually indicates that alternative ways to compare treatments for a response criterion with adjustment for a set of covariables mutually support the same conclusion about the strength of treatment effects. This robustness is noteworthy since the alternative methods for covariance analysis have substantially different rationales and assumptions. Since findings can differ in important ways across alternative choices for covariables (as opposed to methods for covariance adjustment), the critical consideration for studies with covariance analyses planned as the primary method for comparing treatments is the specification of the covariables in the protocol (or in an amendment or formal plan prior to any unmasking of the study.     

A quasi-experimental approach to assessing treatment effects in the nonequivalent control group design.

Evaluates 4 statistical tests of treatment effect for the nonequivalent control group design. This design consists of pre- and posttreatment measures of a dependent variable with biased assignments to treatment groups. The biased assignment creates a treatment-pretest confounding for which different statistical techniques adjust. The different statistical tests discussed are the analysis of covariance, analysis of covariance with reliability correction, raw change score analysis, and standardized change score analysis. If assignment to treatment groups is based on the pretest score (a very infrequent event), analysis of covariance is the appropriate mode of analysis. Selection based on the pretest true scores necessitates a reliability correction procedure. Selection based on stable group differences and selection that occurs midway between the pre- and posttest necessitates change score analysis. (35 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Control of spurious association and the reliability of the controlled variable.

The techniques of matched groups, analysis of covariance, and partial correlation represent various approaches to the prevention of a spurious association between X1 and X2 due to a confounding variable, X3. In all these techniques the use of an unreliable measure for X3 leads to a systematic bias of undercorrection. Adequate corrections are possible for the case of known reliability of X3. Groups should be matched on true scores rather than observed scores, but no correction is possible for the factorial design in which groups are formed on the basis of unreliable correlated measures. Partial correlations should be corrected for the effects of unreliability of the controlled variable. Spuriously high partials are usually obtained when this correction is not applied. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Nonrandom assignment and the analysis of covariance.

Discusses the appropriate use of the analysis of covariance for cases in which groups differ substantially on a variable that is entered as a covariate. The erroneous notions that groups must not differ significantly on the covariate and that covariates must be measured without error are rejected. Selective nonrandom assignment of Ss to groups on the basis of an observed variable that is measured with error can result in groups that differ substantially, but it is shown that conventional analysis of covariance provides unbiased estimates of the true treatment effects, in spite of the initial group differences. In other cases, correction for attenuation due to measurement error is required to obtain unbiased estimates of true treatment effects. (19 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Random versus nonrandom assignment in controlled experiments: Do you get the same answer?

Psychotherapy meta-analyses commonly combine results from controlled experiments that use random and nonrandom assignment without examining whether the 2 methods give the same answer. Results from this article call this practice into question. With the use of outcome studies of marital and family therapy, 64 experiments using random assignment yielded consistently higher mean posttest effects and less variable posttest effects than 36 studies using nonrandom assignment. This difference was reduced by about half by taking into account various covariates, especially pretest effect size levels and various characteristics of control groups. The importance of this finding depends on (a) whether one is discussing meta-analysis or primary experiments, (b) how precise an answer is desired, and (c) whether some adjustment to the data from studies using nonrandom assignment is possible. It is concluded that studies using nonrandom assignment may produce acceptable approximations to results from randomized experiments under some circumstances but that reliance on results from randomized experiments as the gold standard is still well founded. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group

In observational studies, investigators have no control over the treatment assignment. The treated and non-treated (that is, control) groups may have large differences on their observed covariates, and these differences can lead to biased estimates of treatment effects. Even traditional covariance analysis adjustments may be inadequate to eliminate this bias. The propensity score, defined as the conditional probability of being treated given the covariates, can be used to balance the covariates in the two groups, and therefore reduce this bias. In order to estimate the propensity score, one must model the distribution of the treatment indicator variable given the observed covariates. Once estimated the propensity score can be used to reduce bias through matching, stratification (subclassification), regression adjustment, or some combination of all three. In this tutorial we discuss the uses of propensity score methods for bias reduction, give references to the literature and illustrate the uses through applied examples.     

A note on analysis of covariance.

S. H. Evans and E. J. Anastasio (see record 1968-09688-001) constructed 2 sets of hypothetical data to show ways in which the analysis of covariance can yield erroneous results. It is argued that analysis of covariance does not give incorrect results for either set of their data. Analysis of covariance correctly (within rounding error) recovered the form of the models used to construct their data and also provided accurate estimates of parameters for both sets of data. Although analysis of covariance produces accurate results, it must be used cautiously because it may or may not answer the substantive question a researcher asks. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Errata: "Effects of Client Preference and Expectancy upon the Initial Interview".

Reports an error in the original article "Effects of Client Preference and Expectancy upon the Initial Interview" by Andris Ziemelis (Journal of Counseling Psychology, 1974[Jan], 21[1], 23-30). On page 27, an error occurred in Table 1. The F value in the Expectancy column for the Client Satisfaction variable should read 1.19. (The following abstract of this article originally appeared in record 1974-26107-001.) Used a 2×3 design with 10 Ss per cell to show how initial counseling interview process and outcome are affected by (a) assigning clients to either more preferred or less preferred counselors; (b) giving clients either positive, nonexistent, or negative expectancy inductions regarding their counselor assignments; and (c) congruence and incongruence between clients' expectations and experiences. Process and outcome were assessed using client and counselor self-report measures and audiotape segment ratings. More preferred assignment conditions received more favorable tape ratings than did less preferred assignment conditions. Positive or no-expectancy conditions revealed a more favorable counseling outcome than did negative expectancy conditions. Most Ss showed increased preferences for counselors seen despite congruence or incongruence between their expectations and experiences. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Pupillographic assessment of sleepiness in sleep-deprived healthy subjects

Medical research frequently involves the statistical comparison of >2 groups, often using data obtained through the application of complex experimental designs. Fortunately, inferential statistical methodologies exist to address these situations. Analysis of variance (ANOVA) in its many forms is used to simultaneously test the equality of all groups in a study. One-way (with 1 independent variable), 2-way (with 2 independent variables), and repeated-measures (patients serve as their own controls) ANOVAs are forms of this technique. Each form has been developed to analyze data from a specific experimental design. Analysis of covariance (ANCOVA) allows the researcher to control for confounding variables that may influence the response of the dependent variable. Finally, multivariate analysis of variance (MANOVA) evaluates the simultaneous responses of multiple dependent variables to > or = 1 independent variable. Whereas ANOVA is the correct alternative to statistically inappropriate multiple t-tests, MANOVA is the correct alternative to statistically inappropriate multiple univariate ANOVA calculations. Use of each of these statistical methods requires an appropriate experimental design and data meeting a number of assumptions. When used properly, each of these methods provides a powerful statistical analysis technique.     

The importance of the therapeutic relationship in systematic desensitization.

To determine the importance of therapist warmth in effecting positive behavior change using systematic desensitization, 23 female snake-phobic undergraduates were assigned to 1 of 3 groups: warm therapist, cold therapist, or no-treatment controls. Ss in each group were matched on initial Snake Avoidance Test scores. Posttreatment evaluation on the Snake Avoidance Test occurred after 6 20-min desensitization sessions and follow-up scores were obtained 21/2 mo later. Results show that the warm therapist group improved significantly more than either the cold therapist or control group, with no significant differences between these latter 2 groups. Results suggest that therapist warmth is an important variable in systematic desensitization. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Effects of client preference and expectancy upon the initial interview.

Used a 2 * 3 design with 10 Ss per cell to show how initial counseling interview process and outcome are affected by (a) assigning clients to either more preferred or less preferred counselors; (b) giving clients either positive, nonexistent, or negative expectancy inductions regarding their counselor assignments; and (c) congruence and incongruence between clients' expectations and experiences. Process and outcome were assessed using client and counselor self-report measures and audiotape segment ratings. More preferred assignment conditions received more favorable tape ratings than did less preferred assignment conditions. Positive or no-expectancy conditions revealed a more favorable counseling outcome than did negative expectancy conditions. Most Ss showed increased preferences for counselors seen despite congruence or incongruence between their expectations and experiences. (39 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Computer-assisted attention retraining in head-injured individuals: A controlled efficacy study of an outpatient program.

The efficacy of a computer-assisted attention retraining program was evaluated with 29 outpatients suffering from moderate to severe traumatic brain injury. Ss who were at least 12 mo postinjury were randomly assigned either to the attention training program or a memory training program that served as a control condition. Training lasted 9 wks with 2 2-hr sessions per wk for both groups. The experimental design evaluated outcome by juxtaposing a multiple baseline procedure for a 1st set of measures of attention and memory with a pre- and postgroup comparison that relied on a 2nd set of neuropsychological tests. The experimental group improved significantly in comparison with the control group on measures of attention. The reversed pattern for the memory measures was not observed. None of the treatment effects generalized to the 2nd set of dependent variables. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Penalization versus part-time occlusion and binocular outcome in treatment of strabismic amblyopia

OBJECTIVE: The purpose of the study is to compare the visual outcome of occlusion versus penalization treatment of strabismic amblyopia, with particular attention to binocularity outcome. DESIGN: The study design was a retrospective study. PARTICIPANTS: Patients with strabismic amblyopia, 75 receiving penalization alone, 87 with a history of occlusion treatment who were later treated by penalization, and 30 treated by means of part-time occlusion (2 to 6 hours/day) participated in this study. MAIN OUTCOME MEASURES: Logarithm of the minimum angle of resolution (logMAR) visual acuity and binocularity index were measured. RESULTS: No statistically significant difference was found between outcomes for the penalization groups with and without a history of occlusion, either by univariate analysis or by multivariate analysis controlling for initial-visit age, acuity, and binocularity status. One marginally significant outcome difference was found between the pure penalization and part-time occlusion groups by univariate analysis, but no significant difference was found in the multivariate analyses controlling for the same three variables at the initial visit. All visual outcome differences between the pure penalization and part-time occlusion groups were less than 1 logMAR line visual acuity or less than a half-unit on the binocularity index. CONCLUSIONS: The study provided no evidence of a difference in visual function outcome between penalization and occlusion, in terms of either statistical or clinical significance, although limitations of the patient samples used preclude these data from showing conclusively that there was no such difference. The lack of any other study adequately comparing these two treatment methods, in combination with the current study's demonstration of the difficulty of making adequate retrospective-based comparison despite a large patient base (n = 1413), suggests that a large prospective, randomized comparative treatment trial is needed. If atropine penalization, with its high acceptability to patients and parents, is found to produce results comparable with those of occlusion in cases of mild-to-moderate amblyopia, as the current and previous smaller studies suggest, then reconsideration of the standard of care for such amblyopia cases is indicated.     

Cross-lagged panel correlation: A test for spuriousness.

Cross-lagged panel correlation is a method for testing spuriousness by comparing cross-lagged correlations. True experiments control for spuriousness by random assignment, but random assignment limits true experimental studies to independent variables that can be manipulated. Like any statistical method cross-lagged analysis is based on a set of assumptions: synchronicity and stationarity. Different forms of stationarity have different consequences for both the changes in the synchronous correlations over time and the difference between cross-lags. Homogeneous stability is a necessary assumption in the identification of both the source and direction of a causal effect. Cross-lagged analysis is a low-power test, better adapted than either multiple regression or factor analysis for many questions in panel studies. Multiple regression must assume no errors of measurement in the independent variables and no correlated errors, while factor analysis must specify a particular factor structure. Two extended examples of cross-lagged analysis are discussed with special emphasis placed on the issue of stationarity and the estimation of reliability ratios. (50 ref) (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Cytotoxicity of 1-amino-4-phenyl-1,2,3,6-tetrahydropyridine and 1-amino-4-phenylpyridinium ion, 1-amino analogues of MPTP and MPP+, to clonal pheochromocytoma PC12 cells

AIMS: To compare the tolerability and efficacy of a fixed combination solution of dorzolamide/timolol (Cosopt), administered twice daily with the concomitant administration of its components, dorzolamide (Trusopt) twice daily and timolol (Timoptic) twice daily. METHODS: After a 2 week timolol run in, patients with open angle glaucoma or ocular hypertension were randomised (1:1) to receive treatment with either the dorzolamide/timolol combination solution twice daily (combination) or the dorzolamide solution twice daily plus timolol maleate solution twice daily (concomitant) for 3 months. RESULTS: 299 patients were entered and 290 patients completed the study. Compared with the timolol baseline, additional IOP lowering of 16% was observed at trough (hour 0) and 22% at peak (hour 2) at month 3 in both the concomitant and combination groups. The IOP lowering effects of the two treatment groups were clinically and statistically equivalent as demonstrated by the extremely small point differences (concomitant--combination) observed in this study--0.01 mm Hg at trough and 0.08 mm Hg at peak. The safety variables of the concomitant and combination groups were very similar. Both combination and concomitant therapy were well tolerated and few patients discontinued due to adverse effects. CONCLUSIONS: The dorzolamide/timolol combination solution administered twice daily is equivalent in efficacy and has a similar safety profile to the concomitant administration of the components administered twice daily.     

Bias in the assessment of psychotherapy.

Using an outcome study design, 17 therapists rated 54 male veteran psychiatric outpatients on 6 Interpersonal Behavior Inventory (IBI) factors before and after a 4-mo period of psychotherapy. Therapists were divided into 2 groups and given different emphases to selectively augment any bias on either IBI factor, Mistrust (Mis) or Inhibition (Inh). To obtain these emphases, therapists in 1 group rated each of their Ss 5 extra times during the study on Mis alone; the other group made similar ratings on Inh. Analyses of covariance showed no consistent exaggeration of improvement on emphasized factors. Tentative evidence was found in 1 therapist group that some had exaggerated improvement while others had shown an opposite reverse bias. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Statistical methods for analyses of incomplete clinical service records: Concurrent use of longitudinal and cross-sectional data.

Statistical procedures that have become routine in other social sciences were used to analyze data from clinical service records. Despite the absence of control groups, nonrandom assignment of clients to treatment conditions, and incomplete records, effective analyses of psychotherapeutic processes were possible. Multivariate regression models, with variables that were transformed to significantly improve skewness and regression linearity, were controlled for heteroskedasticity and for end-point censoring of dependent variables. They were also used to measure the effects of a categorical variable (gender) and a scalable variable (intake distress) on a reactive outcome measure (of acute distress) and on an unreactive one (of long-term satisfaction). Graphical methods for summarizing large data sets helped identify intake variables that could control for attrition-related sampling biases. These longitudinal covariates and corrections to adjust degrees of freedom for cases with repeated measures were then used to construct statistical models that were equivalents of pure cross-sectional designs. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Shearing injuries of parasagittal white matter, corpus callosum and basal ganglia: possible radiological evidences of hemiplegia in diffuse axonal injury

The relationship between spastic hemiplegia in diffuse axonal injury (DAI) and neuroradiological findings was studied in 100 cases. These cases were prospectively collected from the files of Automobile Insurance Rating Organization in Japan between 1993 from to 1996. Requirements for entry to this study were as follows: presence of initial unconsciousness after head injury without any lucid interval. Existence of CT scan or MRI film obtained within 12 hours of injury showing no significant mass effects, as well as follow-up CT scan or MRI film obtained more than 3 months after the injury. Psychosocial outcome was described both by the medical professional and the caregiver. The hemiplegia was rated severe, mild, or none. The outcome and diffuse ventriculomegaly were classified as reported by the authors previously. Spastic hemiplegia or quadriplegia was documented in the chronic stage in 63 cases, including 53 severe cases with difficulty in walking and 10 mild cases with only pyramidal signs detected. Chi-square analysis showed significant correlation between hemiplegia and the DAI outcome level or ventriculomegaly rating. Focal brain contusion was noticed in 33 cases, but did not correlate with the hemiplegia at all. Radiological findings included 25 cases of parasagittal white matter injury (gliding contusion), 20 cases of callosal injury, 19 cases of basal ganglionic region injury, 5 cases of brain-stem injury, and 3 cases of cerebellar injury. Chi-square analyses of hemiplegia and contralateral presence of these injuries were significant in the former three types of injury. Presence of at least one of these 3 lesions was defined as GCB injury. There were altogether 46 GCB injury cases which were significantly correlated with contralateral hemiplegia by chi-square analysis and by Spearman rank analysis. Partial correlation analysis with hemiplegia as the target variable indicated highly significant correlation only with GCB injury and outcome level. In conclusion, spastic hemiplegia in DAI is a manifestation of primary shear injury. Neuroradiological findings of GCB injury were statistically able to be significantly correlated with the presence of hemiplegia, and suggested pyramidal tract injury either at the corona radiata or the internal capsule level.     

Modeling injury outcomes using time-to-event methods

BACKGROUND: Mortality is an important measurement of injury outcomes, but measurements reflecting disability or cost are also important. Hospital length of stay (LOS) has been used as an outcome variable, but reduced LOS could be achieved either by improved care or by increased mortality. A solution to this statistical problem of "competing risks" would enable injury outcomes based on LOS to be modeled using time-to-event methods. METHODS: Time-to-event methodology was applied to 2,106 cases with complete data from the 1991-1994 registry of a regional trauma center. LOS was used as the outcome variable, modified by assigning an arbitrarily long LOS to any fatal case. A combination of proportional hazards and logistic regression models was used to explore the effects of potential predictive variables, including Trauma Score (TS), Injury Severity Score (ISS), components of TS or ISS, age, sex, alcohol use, and whether a patient was transferred. RESULTS: The "TRISS" combination of TS, ISS, and age previously shown to predict mortality also predicted "modified LOS" (Wald p value less than 0.001 for each variable). Models using only age and certain components of ISS or TS fit our data even better, with fewer parameters. Other variables were not predictive. Modified Kaplan-Meier plots provided easily interpreted graphical results, combining both mortality and LOS information. CONCLUSIONS: With a simple modification to allow for competing risks, time-to-event methods enable more informative modeling of injury outcomes than binary (lived/died) methods alone. Such models may be useful for describing and comparing groups of hospitalized trauma patients.