Showing papers on "Confidence interval published in 1985"

Exact Confidence Intervals for Heritability on a Progeny Mean Basis1

Abstract: Heritability (H) on a progeny mean basis is frequently estimated in recurrent selection experiments for the purpose of estimating the expected progress from family selection; however, appropriate measures of precision have been developed for only a few heritability estimators. The objective of this research was to develop a measure of precision for H for certain balanced linear models. Exact confidence intervals for H were derived and are not restricted to a specific experimental design. The confidence intervals were applied to sorghum [Sorghum bicolor (L.) Moench] half-sib family data.

Evaluation of survival and cause-specific mortality rates using telemetry data

Abstract: Methods are presented for estimating survival and cause-specific mortality rates from radiomarked animals. Time is partitioned into intervals during which the daily rates are assumed to be constant. The rates are estimated from the number of transmitter-days, the number of mortalities due to particular causes, and the number of days in the time intervals. Potential biases arising from combining data from several individuals marked at different times within an interval or from combining rates from different intervals are identified. Variances and confidence intervals for the estimators are presented. Hypothesis testing and sample-size considerations are also illustrated. Simulation showed that the influence of errors in date of death was small, but misdiagnosis of fate had serious consequences. A microcomputer program is available for performing the analyses. J. WILDL. MANAGE. 49(3):668-674 Important mortality agents for wildlife species are legal hunting, wounding loss, poaching, predation, weather, accidents, and disease. Most mark-and-recapture techniques allow investigators to, at most, partition mortality into hunting and "other" sources (Anderson 1975). Radiotelemetry techniques should enable the importance of cause-specific mortality factors to be determined because tagged animals can be located soon after death and the agent of mortality ascertained, regardless of the cause of death. In most radiotelemetry studies, however, the importance of a mortality factor is given as the number of deaths caused by the agent expressed as a percent (Dumke and Pils 1973, Trent and Rongstad 1974, Brand et al. 1975, Trainer et al. 1981). If a sample of animals is marked at the start of a period of interest, calculation of survival and cause-specific mortality rates as simple percentages is appropriate (Hessler et al. 1970). But more often, animals are radiomarked at more than one time, even during different periods within a year for which survival rates differ, and this may lead to two serious biases. First, when some animals are marked midway through a period, animals that died early are not available for sampling and this biases the observed survival rate upwards. And second, if daily survival rates are not constant between intervals, intervals within the period that have the largest sample sizes are most influential in determining the estimates. Methods that avoid these biases with regard to survival rates were first adapted by wildlife biologists to estimate nesting success (Mayfield 1961, 1975; Miller and Johnson 1978; Johnson 1979; Bart and Robson 1982) and have been This content downloaded from 157.55.39.35 on Tue, 30 Aug 2016 05:07:11 UTC All use subject to http://about.jstor.org/terms J. Wildl. Manage. 49(3):1985 SURVIVAL-MORTALITY RATES * Heisey and Fuller 669 Table 1. Definition of variables used in models to determine survival rates and cause-specific mortality rates from telemetry data. i = Interval number (i = 1, 2, ... , I) j = Cause of mortality (j = 1, 2, . , J) xi = Total number of transmitter-days during interval i yi = Total number of mortalities occurring during interval i yij = Total number of mortalities occurring during interval i due to cause j Li = Total number of days in interval i si = Daily survival rate during interval i Si = Survival rate for entire interval i S* = Survival rate for all I intervals mij = Daily mortality rate in interval i due to source j Mij = Interval mortality rate for entire interval i due to source j Mj* = Mortality rate for all I intervals due to source j applied to survival rates in wildlife radiotelemetry studies (Gilmer et al. 1974, Trent and Rongstad 1974, Brand et al. 1975, Trainer et al. 1981). Agent-specific mortality rates in telemetry studies have been evaluated only as simple percentages that are subject to the aforementioned biases. In this paper, we generalize the basic approach of Mayfield (1961, 1975) and Trent and Rongstad (1974) to determine unbiased estimates of cause-specific mortality rates. Standard statistical methodology for competing risk analysis is used (Chiang 1968, Kalbfleisch and Prentice 1980). Radiotelemetry applications are emphasized, although the techniques are applicable to any situation where the subjects can be relocated at will and the cause of death identified. A microcomputer program called MICROMORT is available which performs most of the described calculations for moderate size problems. Versions are available for the IBM PC. A diskette containing the program and documentation can be requested from T. K. Fuller. We are grateful to J. R. Cary, D. H. Johnson, and G. C. White for their constructive comments and encouragement. E. K. Fritzell, J. M. Hoenig, D. W. Kuehn, and R. E. Lake also made helpful suggestions. Reprint requests should be made to Todd K. Fuller.

Bootstrap confidence intervals for a class of parametric problems

Abstract: SUMMARY We consider the following class of problems: having observed a multivariate normal data vector y with unknown mean vector q, covariance matrix the identity, find an approximate confidence interval for 0 = t(q), a real-valued function of q. A simple geometric construction is given which leads to highly accurate solutions. This construction shows that the standard approximation based on maximum likelihood theory, 04+(CZ00, can be quite misleading when 0 is nonlinear in q. We discuss bootstrap-based confidence intervals which remove most of the error in the standard approximation, at the expense of considerably more calculation. The bootstrap intervals are invariant under transformation of both y and i, and so they automatically produce accurate solutions in problems which can be transformed to multivariate normality, without requiring knowledge of the normalizing transformation.

The MRFIT behavior pattern study. II. Type A behavior and incidence of coronary heart disease.

Abstract: Behavior pattern was assessed by interview for 3,110 men at eight centers in the Multiple Risk Factor Intervention Trial (1973-1976). The Type A pattern was not significantly associated with risk of first major coronary events (coronary death and definite nonfatal myocardial infarction) after a mean follow-up of 7.1 years. Crude relative risks for Types A1-A2 versus X-B were 1.08 in usual care, 0.82 in special intervention, and 0.92 overall. Adjustment for age, blood pressure, cigarette smoking, serum cholesterol, consumption of alcohol, and educational attainment yielded relative risks of 0.99 in usual care, 0.81 in special intervention, and 0.87 overall (95% confidence interval = 0.59-1.28). The Jenkins Activity Survey Type A score, obtained for 12,772 men at all 22 centers, was also not significantly associated with risk of first major coronary events. Overall, crude risks in the lowest (Type B) through highest (Type A) quintiles of the score's distribution were 5.0%, 4.4%, 4.0%, 4.3%, and 4.1%, respectively. The proportional hazards regression coefficient, adjusted for the variables listed above, was -0.006 (95% confidence interval = -0.015-0.003). These results raise questions regarding the robustness of the Type A hypothesis in its present form. Further studies are needed to investigate these questions and to evaluate the validity of procedures used to assess behavior patterns.

Type of Occupation and Near-Future Hospitalization for Myocardial Infarction and Some Other Diagnoses

Abstract: Using three different registers a cohort study was undertaken to describe the relationship between type of occupation and hospitalization. A total of 958 096 subjects aged 20-64 years were followed-up for one year regarding inpatient care. Several significant associations between type of occupation and incidence of hospitalization for different diagnoses were observed. Male subjects employed in occupations where a high proportion reported a combination of hectic work and few possibilities to learn new things were more frequently hospitalized for myocardial infarction than other working men. The relative 'hospitalization' ratio for men 20-54 years of age in these strenuous occupations was estimated as 1.6 with a 95% confidence interval of 1.3-1.9. For women, the relative 'hospitalization' ratio in the ages 20-64 in occupations where a high proportion reported a combination of hectic and monotonous work was estimated as 1.6 with a 95% confidence interval of 1.1-2.3. Both these associations were statistically significant even after controlling for 12 possible confounding factors.

On the Number of Bootstrap Simulations Required to Construct a Confidence Interval

Abstract: : The purpose of this document is to make two points about the effect of the number of bootstrap simulations, B, on percentile-t bootstrap confidence intervals. The first point concerns coverage probability; the second, distance of the simulated critical point form the true critical point derived with B=infinity. In both cases the author has in mind applications to smooth statistics, such as the Studentized mean of a sample drawn from a continuous distribution. He indicates the change that have to be made if the distribution of the statistic is not smooth. Additional keywords: Exponentive functions.

Confidence limits for estimates of gene linkage based on analysis of recombinant inbred strains

Abstract: Recombinant inbred (RI) strains are extremely useful for genetic mapping. This paper presents a simple method for determining confidence intervals for linkage estimates based on analysis of RI strains. The results show that such confidence intervals are usually large with the currently available numbers of RI strains. Therefore, map positions based only on analysis of RI strains should be interpreted with caution. To facilitate interpretation of linkage data derived from RI strains, a table is presented giving the 95 percent and 99 percent confidence intervals for all possible linkages detected with up to 45 RI strains.

When was a negativë clinical trial big enough? how many patients you needed depends on what you found

Abstract: • "Negative" clinical trials that conclude that neither of the treatments is superior are often criticized for having enrolled too few patients. These criticisms usually are based on formal sample size calculations that compute the numbers of patients required prospectively, as if the trial had not yet been carried out. We suggest that this "prospective" sample size calculation is incorrect, for once the trial is over we have "hard" data from which to estimate the actual size of the treatment effect. We can either generate confidence limits around the observed treatment effect or retrospectively compare it with the effect hypothesized before the trial. If the observed effect is small, the risk of the false-negative conclusion (and the sample size required to draw negative or equivalency conclusions) is often much less than that generated by the "prospective" calculation. (Arch Intern Med1985;145:709-712)

Qualms about Bootstrap Confidence Intervals

Abstract: The percentile method and bias-corrected percentile method of Efron (1981, 1982) are discussed. When these methods are used to construct nonparametric confidence intervals for the variance of a normal distribution, the coverage probabilities are substantially below the nominal level for small to moderate samples. This is due to the inapplicability of assumptions underlying the methods. These assumptions are difficult or impossible to check in the complicated situations for which the bootstrap is intended. Therefore, bootstrap confidence intervals should be used with caution in complex problems.

Health experiences of operating room personnel.

Abstract: In an attempt to evaluate health experiences of operating room personnel using previously published reports, the authors calculated summary relative risks (RRs) for each outcome under investigation by combining data from six studies For each summary RR, they also calculated 95% confidence limits; when the range of the confidence interval excludes 10, the increased risk is statistically significant at the 005 level The most consistent evidence was for spontaneous abortion among pregnant physicians and nurses who work in operating rooms, where the RR was 13 (95% confidence limits from 12 to 14) For liver disease there were statistically significant increased RRs among both men (16, 13-19) and women (15, 12-19), but these were based on smaller numbers of studies Although the results of pooled analyses are suggestive, most studies of this issue have relied on voluntary responses and self-reported outcomes, so that response and/or recall bias could explain these findings In addition, these investigations generally have examined working in operating rooms rather than actual exposure to anesthetic gases Finally, there have been considerable improvements in operating room scavenging systems during the last decade Thus, prospective cohort studies are needed to determine whether there is a relationship between current levels of occupational exposure to anesthetic gases and adverse outcomes, particularly spontaneous abortion and liver disease

Age at first primary as a determinant of the incidence of bilateral breast cancer. Cumulative and relative risks in a population-based case-control study.

Abstract: This epidemiologic investigation comprised 1351 of 1423 women in a defined geographic area consecutively diagnosed as having a primary breast cancer. Simultaneous bilateral disease occurred in only 1 patient, whereas a history of previous cancer in the contralateral breast was reported by 65 patients. This prevalence was related to that of 23 previous cases in an age-matched control group of 1351 women from the same population. The relative risk of developing a second primary was 2.9 (95% confidence limit, 1.8-4.6) for the whole material and remained seemingly constant over several decades at a level predetermined by age at first diagnosis, namely 9.9 (95% confidence limit, 3.8-25.8) before the age of 50 and 1.9 (95% confidence limit, 1.1-3.2) after that age. The incidence ratio of bilateral to unilateral disease was used as an estimate of the lifetime risk of developing a second primary in this stable and well-defined population. This calculation revealed cumulative risk figures of 13.3% and 3.5% for women younger and older than 50 years, respectively, at first diagnosis. It was concluded that the occurrence of bilateral disease, which reflects a multicentric neoplastic transformation of the breast epithelium, is a characteristic of early-occurring (premenopausal) disease.

Variance Components Analysis: A Selective Literature Survey

Abstract: Summary A survey is given of developments in the area of variance components during the last three decades. The survey covers mainly point estimation, interval estimation, and hypothesis testing concerning the variance components for both balanced and unbalanced models.

Confidence limits for probability of response in multistage phase II clinical trials.

Abstract: Herson (1979, Biometrics 35, 775-783) has given a method for designing one-arm k-stage phase II clinical trials, which permits early termination of the trial if the treatment is apparently ineffective, while retaining acceptable levels of power and significance. This paper gives a method for calculating the confidence limits on the response proportion, conditional on a particular study design.

A nonlinear multiple regression program, MULTI2 (BAYES), based on Bayesian algorithm for microcomputers.

Abstract: A nonlinear multiple regression analysis program MULTI2(BAYES) was developed for microcomputers The Bayesian algorithm which is incorporated in MULTI2 (BAYES) combines the insufficient individual patient data (individual data) with the pharmacokinetic parameters published in literatures (population parameters) to predict the plasma time course of the patient The program is written in the minimum Microsoft BASIC commands alone to be executable on many personal computers without any modification The numbers of parameters to estimate, independent variables and dependent variables are not restricted in use of MULTI2(BAYES) The pharmacokinetic models are defined as one pleases by the user The four nonlinear least squares algorithms, ie Gauss-Newton method, damping Gauss-Newton method, modified Marquardt method by Fletcher and simplex method can be selected at user's option MULTI2(BAYES) calculates the confidence limits of time courses at 95% significant level

Statistical Inference on Measures of Niche Overlap

Abstract: Author(s): Mueller, LD; Altenberg, L | Abstract: Estimates of measures of niche overlap are often reported without any indication of sampling variance or an accompanying confidence interval. The delta, jackknife and bootstrap methods for making statistical inferences on 4 measures of niche overlap were investigated: the coefficient of community, Morisita's index, Horn's index, and the Euclidian distance. Qualitative conclusions are: 1) The bias of these estimators was usually l10% of the mean unless the sample size was small and the number of resource categories large. The jackknife and bootstrap can significantly reduce this bias. 2) The variance of the bootstrap and jackknife estimators was usually greater than that of the 'standard' estimator. 3) Under a variety of circumstances, the population sampled may actually represent several unrecognized subpopulations. In such cases confidence intervals generated by the jackknife and delta techniques can be quite inaccurate, while the nonparametric confidence intervals derived from the bootstrap are highly accurate.-Authors

Confidence intervals for a variance ratio, or for heritability, in an unbalanced mixed linear model

Abstract: A procedure is presented for constructing an exact confidence interval for the ratio of the two variance components in a possibly unbalanced mixed linear model that contains a single set of m random effects. This procedure can be used in animal and plant breeding problems to obtain an exact confidence interval for a heritability. The confidence interval can be defined in terms of the output of a least squares analysis. It can be computed by a graphical or iterative technique requiring the diagonalization of an m X m matrix or, alternatively, the inversion of a number of m X m matrices. Confidence intervals that are approximate can be obtained with much less computational burden, using either of two approaches. The various confidence interval procedures can be extended to some problems in which the mixed linear model contains more than one set of random effects. Corresponding to each interval procedure is a significance test and one or more estimators.

Do Oral Contraceptives Prevent Rheumatoid Arthritis

Abstract: Two studies have suggested that the risk of rheumatoid arthritis in women using oral contraceptives is less than half that of nonusers. When a third study from the Mayo Clinic failed to confirm these findings, it was criticized for inclusion of ineligible subjects, misclassification of oral contraceptive use, and inadequate statistical power. Recent expansion of the Mayo Clinic's data resources provided a unique opportunity to resolve the controversy, and a new population-based case-control study was undertaken. In comparison with the previous study, the new investigation had 2.2 times as many eligible cases and more complete ascertainment of oral contraceptive use via access to the records of Planned Parenthood of Minnesota. Comparing any prior use of oral contraceptives with never having used them, the relative risk of rheumatoid arthritis estimated from 182 cases and their 182 matched controls was 1.1 (95% confidence interval, 0.7 to 1.7). The relative risk for current use was 1.3 (95% confidence interval, 0.7 to 2.4). The lack of a protective effect was independent of age, disease severity, and disease end point (date of confirmed diagnosis or symptom onset). ( JAMA 1985;254:1938-1941)

Back pain and heavy physical work: a comparative study of concrete reinforcement workers and maintenance house painters.

Abstract: In an investigation of the effect of heavy physical work on the back 217 concrete reinforcement workers aged 25-54 and a reference group of 202 house painters of similar age were interviewed about their back symptoms. Data on occupational history, accidents, and leisure time activities were collected with a questionnaire. The cumulative incidence rate of sciatic pain was significantly higher among the reinforcement workers than the painters. As regards the occurrence of lumbago and non-specific back pain, however, the groups were alike. In both occupational groups sciatic pain during the previous 12 months was associated with earlier back accidents (odds ratio 2.8, 95% confidence interval 1.8-4.5). The accident rate of the reinforcement workers was higher than that of the painters, and this difference seemed to explain their higher rate of sciatic pain.

The effect of subsampling sites within patients.

Abstract: In recent periodontal research, multiple sites within patients have been utilized as the primary observations or sampling units. In the present study a simple approach for the single sample case was used to demonstrate the effect of site-specific analyses when the patient is not treated as the sampling unit. This approach was shown to underestimate the variance of the mean value in question and thus, in tests of hypotheses concerning the mean, lowers artificially the p-level or equivalently raises artificially the 1-α confidence level of intervals. The error introduced by this procedure increased when the number of actual primary random sampling units (the patients) was small, for this resulted in inappropriately lowered Student's tvalues. The error also increased as the number of sites used within patients increased and decreased slightly as the number of patients increased. From only six periodontal patients where six randomly chosen sites were observed for pocket depth in mm, a statistically significant (p<0.01) intraclass correlation was seen (ϱ1=0.31). If the 36 sites had been treated as independently distributed, a true probability of p(true)=0.27 would have been misread nominally as p(nominal)=0.05.

Prevalence of dental mottling in school-aged lifetime residents of 16 Texas communities.

Abstract: The severity of dental mottling in 2,592 school-aged, lifetime residents of 16 Texas communities was investigated in 1980-81 to identify factors associated with mottling and to construct a prediction model for the prevalence of mottling. The communities were selected to obtain a wide range of levels of fluoride in the drinking water. The children within each of the communities were contacted through their schools and received a dental examination to assess the severity of mottling. Information on demographic, dental health practice, and other candidate predictor variables was obtained from a questionnaire completed by a parent. A number of water quality measurements were also recorded for each community. White and Spanish-surname children had about the same prevalence of mottling while Blacks had a higher prevalence, odds ratio (OR) = 2.3, 95% confidence interval = 1.4, 3.7. Children from homes which had air conditioning had a lower prevalence of mottling (OR = .6, (0.4, 0.8)). The use of fluoride toothpaste or drops and the number of fluoride treatments were almost identical among those who did and did not develop moderate mottling. In addition to fluoride, total dissolved solids and zinc were water quality variables associated with mottling.

Cigarette smoking and Down syndrome.

Abstract: A matched case-control study of 100 mothers of Down syndrome children, 100 mothers of children with other defects (defect controls), and 100 mothers of children with no defects (normal controls) was carried out. All infants were born in upstate New York in 1980 and 1981. Matching was very close on maternal age for the normal controls but not for the defect controls. The risk ratios for the association of cigarette smoking around the time of conception with Down syndrome was 0.58 (90% confidence interval of 0.34-0.98) in the case-defect control comparison and 0.56 (90% confidence interval of 0.33-0.95) in the case-normal control comparison. Stratification by alcohol ingestion and maternal age did not abolish the negative trend to association. The results are contrary to that of an earlier study of others that found a positive association of older age and trisomy in spontaneous abortions. In fact, among mothers of Down syndrome cases over age 30 in this analysis, the risk ratio was lower than for younger mothers. (For case-normal control comparisons, the value was 0.39 [90% confidence interval of 0.17-0.87]). If not due to chance or confounding, the negative association in our data may be attributable to, among other factors, a selective effect of smoking upon survival or fertilizability of +21 gametes prior to conception or upon survival of +21 conceptuses after fertilization.

A Case-Control Study of BCG and Childhood Tuberculosis in Cali, Colombia

Abstract: We conducted a case-control study to evaluate the effectiveness of BCG vaccination in preventing childhood tuberculosis (TB) in Cali, Colombia. We ascertained 178 cases aged 0 to 14 years from the respiratory clinics with cough or fever for at least three weeks and a positive chest X-ray for TB, as well as 320 controls who were from the same households but had no symptoms and negative X-rays. Using matched set multiple logistic regression analysis, we found the age- and sex-adjusted relative risk (RR) of TB among vaccinees compared with non-vaccinees to be 0.84 with 95% confidence limits (CL) from 0.43 to 1.62. There was, however, a significantly lowered relative risk of TB with increasing time since vaccination (RR = 0.83 per year since time of vaccination with 95% CL from 0.74 to 0.94.)

Repeated confidence intervals for the median survival time

Abstract: SUMMARY We describe methods for the construction of a confidence interval for median survival time based on right-censored data. These methods are extended to the construction of repeated confidence intervals for the median, based on accumulating data; here, the overall probability that all intervals contain the true median is guaranteed at a fixed level. The use of repeated confidence intervals for median survival time in post- marketing surveillance is discussed. A confidence interval for median survival time provides a useful summary of the survival experience of a group of patients. If confidence intervals are calculated repeatedly, as data accumulates, the probability that at least one interval fails to contain the median may be much higher than the error rate for a single interval, and if these confidence intervals are used in a decision making process the probability of an incorrect decision increases accordingly. Jennison & Turnbull (1984) have proposed methods for calculating repeated confidence intervals appropriate to such situations. Similar ideas have also been discussed by Lai (1984). In ? 2 we propose a new form of single-sample nonparametric confidence interval; this interval has asymptotically correct coverage probability and Monte Carlo simulations suggest it is superior to its competitors for small sample sizes. Repeated confidence intervals for the median are presented in ? 3 and their stnall sample size performance is assessed by Monte Carlo simulation; an example of their use is given in ? 4. All the methods considered can easily be modified to give confidence intervals for other quantiles or for the survival probability at a fixed time.

Statistical evaluation of the jackknife estimate of diversity when using quadrat samples

Abstract: The jackknife estimation procedure is used to estimate Brillouin's and Simpson's indices of diversity when sampling is done by selecting quadrats. Using simulated data, a sampling plan is set up to study effects of quadrat size, sample size, and sampling area. The estimates are evaluated by estimating bias, standard error, and mean square error, and by comparing expected 95% confidence interval coverage with observed 95% confidence interval coverage. Smaller quadrats tend to give less biased and smaller variance estimates of Brillouin's index. The jackknife estimate of Simpson's index is unbiased using quadrat samples, but the estimates of standard error appears to be an overestimate for large samples, causing overcoverage (>95%) for observed confidence intervals. See full-text article at JSTOR

A new variance estimator for the Mantel-Haenszel odds ratio

Abstract: Many asymptotic formulas for estimating the variance of and confidence intervals for odds ratios in case-control studies are available, but most have limitations. The method suggested by Cornfield (1956, Proceedings of the Third Berkeley Symposium 4, 135-148), for example, is iterative and requires calculation of exact variances if the number of subjects within each stratum is small. This paper presents a new asymptotic, noniterative variance estimate for the Mantel-Haenszel odds ratio which, based on a series of Monte Carlo experiments, compares favorably with other noniterative estimates when used for confidence interval estimation.