Prevention is better than cureGiven the above considerations, plagiarism is better dealt with by concentrating on an instructional system of prevention.Many institutions have begun to implement courses on the responsible conduct of research, exploring a range of research integrity issues.Although this is a step in the right direction, such instruction benefits mainly students.Legions of seasoned researchers continue to operate without such instruction.Clearly, research integrity needs to be incorporated into continuing education targeted at this group.Training on responsible conduct of research must go beyond the "high crimes" of fabrication, falsification, and plagiarism, all three of which are thankfully still relatively rare in science.Zigmond and Fischer have argued that misdemeanours make up the bulk of the ethical transgressions. 10Instruction on plagiarism should focus on the principles of ethical writing. 11This approach assumes that each of our written works represents an implicit contract between us and our readers in which the reader assumes that, unless otherwise noted, we are the sole authors of the work, the words and ideas are our own, and the ideas, concepts and theories described are accurately and objectively represented to the best of our ability.Since it is doubtful that we will ever eliminate plagiarism or other forms of intentional misconduct, we need to increase our ability to detect these transgressions and effectively prosecute and punish offenders.Instruction in ethical writing indirectly touches on several other traditional forms of misconduct.Consequently, I believe that when we internalise and apply its basic principles they will significantly reduce plagiarism and generalise to other areas of scientific research and personal conduct.I thank Maryellen Reardon for comments on an earlier draft of this article.

OBJECTIVE:To evaluate the impact of an interventional multidisciplinary antibiotic management program on expenditures for antibiotics and on the incidence of nosocomial infections caused by Clostridium difficile and antibiotic-resistant pathogens during 7 years. DESIGN:Prospective study with comparison with preintervention trends. SETTING:University-affiliated teaching hospital. PATIENTS:All adult inpatients. INTERVENTION:A multidisciplinary antibiotic management program to minimize the inappropriate use of third-generation cephalosporins was implemented in 1991. Its impact was evaluated prospectively. The incidence of nosocomial C. difficile and resistant Enterobacteriaceae infections as well as the rate of vancomycin-resistant enterococci (VRE) and methicillin-resistant Staphylococcus aureus (MRSA) were compared with those of National Nosocomial Infections Surveillance System hospitals of similar size. RESULTS:Following implementation of the program, there was a 22% decrease in the use of parenteral broad-spectrum antibiotics (P < .0001) despite a 15% increase in acuity of patient care during the following 7 years. Concomitantly, there was a significant (P = .002) decrease in nosocomial infections caused by C. difficile and a significant (P = .02) decrease in nosocomial infections caused by resistant Enterobacteriaceae. The program also appeared to have a favorable impact on VRE rates without a sustained impact on MRSA rates. CONCLUSION:These results suggest that an ongoing multidisciplinary antibiotic management program may have a sustained beneficial impact on both expenditures for antibiotics and the incidence of nosocomial infection by C. difficile and resistant bacterial pathogens.

Let (X t ,t∈Z) be a linear sequence with non-Gaussian innovations and a spectral density which varies regularly at low frequencies. This includes situations, known as strong (or long-range) dependence, where the spectral density diverges at the origin. We study quadratic forms of bivariate Appell polynomials of the sequence (X t ) and provide general conditions for these quadratic forms, adequately normalized, to converge to a non-Gaussian distribution. We consider, in particular, circumstances where strong and weak dependence interact. The limit is expressed in terms of multiple Wiener-Itô integrals involving correlated Gaussian measures.

Abstract.Long memory is known to occur in many fields of statistical application. Stationary processes whose correlations decay asymptotically like ‖k‖2H‐2, wherekis the lag andHε (0.5, 1), provide useful parsimonious models with long memory. The parameterHcharacterizes the long‐memory features of the data. For long time series, maximum likelihood estimation ofHcan be costly in terms of CPU time. In this paper, we show that, for disjoint stretches of the data, estimates ofHand other parameters that characterize the dependence structure are asymptotically independent. Averaging these estimates leads to a fast and efficient approximate maximum likelihood method.

We consider a general long memory time series, assumed stationary and linear, but not necessarily Gaussian or generated by a finite-parameter model. For such a process, we derive the asymptotic joint distribution of the normalized periodogram at a fixed, finite collection of Fourier frequencies. The limiting distribution is represented in terms of Wiener-Ito integrals, and, for a single periodogram ordinate, it is an unequally weighted linear combination of independent chi(1)(2) random variables. This result was previously known only in the Gaussian case. Our theorem may be useful for generalizing, beyond the Gaussian case, the applicability of a semiparametric method of estimating the long memory parameter based on log-periodogram regression.

Consider integrals whose integrands are products of functions bounded near zero by powers of affine functionals and near infinity by different powers of the same functionals. The power counting theorem provides conditions for integrability.

Normalized quadratic forms of moving averages converge to double Wiener-Itô integrals if the summands are sufficiently dependent. This result extends to sums of bivariate Appell polynomials of arbitrary degree.

Certain quadratic forms with long-range dependence, normalized by Nd with d > 12, have a non-Gaussian limit, but under further normalization, as d → 12, the limit becomes Gaussian.

We examine the limit behavior of quadratic forms of stationary Gaussian sequences with long-range dependence. The matrix that characterizes the quadratic form is Toeplitz and the Fourier transform of its entries is a regularly varying function at the origin. The spectral density of the stationary sequence is also regularly varying at the origin. We show that the normalized quadratic form converges inD[0, 1] to a new type of non-Gaussian self-similar process, which can be represented as a Wiener-Itô integral onR2.