Konanur G. Janardan

Grab versus composite sampling: A primer for the manager and engineer

Effluent subsamples are usually aggregated into flow or time proportional samples before analysis. Although this provides information on average process conditions, that on process variability is lost by compositing. Fishers information is defined and used to estimate the loss due to compositing. The results of simulations based on parameters derived from actual waste streams support the fact that random grabs serve as well as composite samples for monitoring purposes. These findings favor changes in regulatory practice to allow compliance to be demonstrated by grab sample averages. Reporting requirements based on moving averages are shown to be inferior to those based on averages taken over nonoverlapping time periods.

Monitoring toxics by group testing

Monitoring the environment for a large list of organic compounds present at low levels is costly. The primary purpose of such monitoring is to assure that hazardous levels of such compounds are not released into, or present in, the environment. Viewing this as a quality control problem, we suggest that samples from different sources can be composited using group testing procedures prior to analysis. Our purpose is to describe the basic concepts and suggest problems requiring study.

Efficiencies of liquid-liquid extraction and XAD-4 and XAD-7 resins in collecting organic compounds from a coke plant's effluent

ConclusionThese results, taken together with our earlier study (Janardan et al. 1980), suggest that the chemical characterization of a complex wastes requires the simultaneous collection of samples by several methods.

Graphical effluent quality control for compliance monitoring: What is a violation?

Demonstration of compliance with discharge (effluent) requirements suffers because Discharge Monitoring Reports (DMRs) give neither the discharger nor the regulator information on the process mean which ensures compliance. To obtain this information, a simple graphical method is presented which takes into account process variability. This technique provides quality control specifications for the monitoring data and the process itself.

Threshold estimation from the linear dose-response model: Method and radiation data

The linear dose-response model is considered a conservative, nonthreshold relationship. This is based on a confusion between the sufficient condition (that is, zero slope at zero dose) and the necessary condition (that is, response distinguishable from zero). Once the threshold is properly defined, it is shown that the linear model predicts thresholds for radiation data in good agreement with experimental results.

Threshold and dose response estimation using the “filter model”

The “filter model” has been developed to explain the biologic effects of radiation and chemicals. We have examined nearly 300 sets of dose response data, of which 50 are presented here. Responses (induced by radiation and chemicals) which have been examined include in vitro survival studies on animal and plant tissues, induction of cellular aberrations and time to tumor or death. Similar data from in vivo studies has also been examined. All of the data appear to fit the model R = a lnD + b(lnD)2 + c, where R is the response, a and b are parameters fitted by regression to a particular set of data, and c is the response at zero (or lowest) dose. By writing this model in exponential form, it can be seen that the response R results from multistage filtering (by net amounts a and b) of the initial dose, D. The threshold is obtained from this model as the point, DT, at which the second derivative becomes zero. This is given by DT = exp(1 − a2b) when a and b are oppositelt signed.

Estimating the proportion of mutagenic compounds in environmental samples.

A method for estimating the proportion of mutagens in a sample of N compounds is developed. For this procedure to be applicable, there must be a statistically significant correlation between the number of mutagens in the sample and the sample size N. Sample size is treated as a random variable. A sequential sampling scheme is considered. In the first stage, compounds are identified and classified as mutagens, nonmutagens, or untested, as reported in the literature. In the second stage, all untested compounds are tested for mutagenicity. Since data of this type are not generally available, estimates of the proportions of compounds tested (p), tested and mutagenic (p1), and untested but mutagenic (p2) are developed from existing complications. It is shown that there is a high, statistically significant correlation between the total number of mutagens in a sample and the sample size N. The proportion of mutagens in a sample for various values of p, p1, and p2 is tabulated.

A new one-parameter model for the distribution of chromosome breaks in human cells.

Damage to genetic material can be brought about by chemicals, radioactivity, radiation, etc. This damage can be visualizedas chromosome aberrations such as chromosome and chromatid deletions, dicentrics, chromatid exchanges, and others. The study of chromosome aberration induction by chemical and physical agents has evolved over several decades and discrete distribution models have been developed to analyze the data generated in experimental studies. Most of these models have relied on the Poisson, binomial or negative binomial distributions (SAVAGE 1970, KOHLER et al. 1976, YANKOVENKO et al. 1976) to describe the frequency distribution of aberrations in a collection of cells. The intracellular distribution of lesions gives useful information about the mechanism of the interaction of the damaging agent with different cell components. Recently, JANARDAN and SCHAEFFER (1977), JANARDAN et al. (1979), and DuFRAIN et al. (1980) have proposed the use of the Langrangian Poisson Distribution (LPD) as a model of over-, under- or no dispersion relative to the Poisson. (The value of ~2, the parameter of the LPD, is a quantitative measure of the dispersion while its sign (+, -, O) indicates the direction.) However, certain asumptions which are used in the development of the LPD are restrictive of the more complex biologic phenomena being studied by cytogeneticists. Further, many workers have questioned the biologic significance of the improvement resulting from the use of two (or more) parameter models over the use of one parameter models, since additional parameters will result in mathematical improvement at the expense of a reduced number of degrees of freedom for the experimental data. In this paper we develop a new one parameter model, which we call the Poisson-Linear Exponential Distribution (PLED). This distribution is as good as many of the two parameter models at fitting experimental data. The assumptions of the model are developed from biologic observations. HYPOTHESES: I. There is a constant rate of aberration for a given cell. 2. The rate of aberrations varies from cell to cell.

Is there a correlation between the number of mutagens and the number of nonmutagens in an environmental sample

Abstract We have adapted a procedure which was developed for studying the correlations in the distribution of genetic factors, such as sex ratios in siblings, to obtain estimates of the correlation between the number of mutagens and the number of nonmutagens in a sample. Positive correlations with correlation coefficients in excess of 0.8 were obtained. The high correlation suggests that it is possible to estimate the number of mutagens in samples with N > 100 compounds as (0.064 ± 0.01) N .