Estimation of Parameters in the Distributions

Most real world models often require multiparameter solutions. So, to characterize a normal distribution an estimate of the mean p and variance a2 is needed. In some instances, only a subset of parameters in the model may be of interest, e.g., only the mean of the distribution may be of interest, in which case the other parameters are considered nuisance parameters. In fitting a likelihood function, however, maximization is done relative to all model parameters. But, if only a subset of model parameters are of interest, a method needs to be available to concentrate the likelihood on the model parameters of interest and eliminate the nuisance parameters. 

Model selection should be based not solely on goodness of fit but also on the degree of confidence of the predicted parameters. It is well known that increasing the number of free parameters to be estimated can improve the goodness of fit but can also decrease the confidence in the estimates of the model parameters. Therefore, ranking the models based solely on standard deviation (SD) data may not be satisfactory for comparing functions with different number of parameters and different sample sizes. In this work, AIC and BIC were used to take into account the different number of parameters in the probability distribution functions when ranking the best functions to use to describe distillation data. 

Several properties of the filler are important to the compounder (279). Properties that are frequentiy reported by fumed sihca manufacturers include the acidity of the filler, nitrogen adsorption, oil absorption, and particle size distribution (280,281). The adsorption techniques provide a measure of the surface area of the filler, whereas oil absorption is an indication of the stmcture of the filler (282). Measurement of the sdanol concentration is critical, and some techniques that are commonly used in the industry to estimate this parameter are the methyl red absorption and methanol wettabihty (273,274,277) tests. Other techniques include various spectroscopies, such as diffuse reflectance infrared spectroscopy (drift), inverse gas chromatography (igc), photoacoustic ir, nmr, Raman, and surface forces apparatus (277,283—290). 

If a measurement is repeated only a few times, the estimate for the distribution variance calculated from this sample is uncertain and the tiornial distribution cannot be applied. In this case another distribution is used, f his distribution is Student s distribution or the /-distribution, and it has one more parameter the number of degrees of freedom, t>. The /-distribution takes into account, through the p parameter, the uncertainty of the variance. The values of the cumulative /-distribution function cannot be evaluated by elementary methods, and tabulated values or other calculation methods have to be used. 

Making a detailed estimate of the full loading of an object by a blast wave is only possible by use of multidimensional gas-dynamic codes such as BLAST (Van den Berg 1990). However, if the problem is sufficiently simplified, analytic methods may do as well. For such methods, it is sufficient to describe the blast wave somewhere in the field in terms of the side-on peak overpressure and the positive-phase duration. Blast models used for vapor cloud explosion blast modeling (Section 4.3) give the distribution of these blast parameters in the explosion s vicinity. 

Even if we make the stringent assumption that errors in the measurement of each variable ( >,. , M.2,...,N, j=l,2,...,R) are independently and identically distributed (i.i.d.) normally with zero mean and constant variance, it is rather difficult to establish the exact distribution of the error term e, in Equation 2.35. This is particularly true when the expression is highly nonlinear. For example, this situation arises in the estimation of parameters for nonlinear thermodynamic models and in the treatment of potentiometric titration data (Sutton and MacGregor. 1977 Sachs. 1976 Englezos et al., 1990a, 1990b). 

To facilitate interpretation of the outputs, the authors also created two simulation data sets with identical distributional properties (number of indicators, number of levels, indicator intercorrelations, skew and kurtosis) one taxonic set and one dimensional set. The taxonic data set was created to have a base rate of. 23, which corresponds to the proportion of cases falling at or above a BDI threshold of 10 in the undergraduate data set. Ruscio and Ruscio tried to ensure that indicator validities and nuisance correlations matched the estimated parameters of the real indicators, but they did not indicate how successful this was. 

The formal statistical comparison of residue distributions across the various subpopulatlons involves estimation of the parameters In the model given by Equation 3. The model assumes that each subpopulation distribution is lognormal but possibly differ in mean residue levels and variances. Significant differences In Census Divisions correspond to significant differences In the Di values. Differences in the Aj s correspond to differences among the age categories. The coefficients Si and 2 provide information concerning race and sex differences. 

In the problem of selecting a distribution for a ID model of variation, there are 2 kinds of variables, namely, 1) the data, which we know and 2) distribution parameters, which will be assigned values based on the data. Here we will often follow statistical terminology by using the term estimation (of parameters) instead of fitting. In statistical terminology, the values assigned to distribution parameters are termed estimates the expressions used to compute estimates are estimators. ... 

Bias corrections are sometimes applied to MLEs (which often have some bias) or other estimates (as explained in the following section, [mean] bias occurs when the mean of the sampling distribution does not equal the parameter to be estimated). A simple bootstrap approach can be used to correct the bias of any estimate (Efron and Tibshirani 1993). A particularly important situation where it is not conventional to use the true MLE is in estimating the variance of a normal distribution. The conventional formula for the sample variance can be written as = SSR/(n - 1) where SSR denotes the sum of squared residuals (observed values, minus mean value) is an unbiased estimator of the variance, whether the data are from a normal distribution... 

One often encounters a distinction between precision and accuracy. Accuracy relates to systematic deviation between parameter estimates and actual parameter values precision relates to the spread in the distribution of estimates. This terminology is not often used explicitly in the estimation theory literature, but the concepts are often implicit. 

With small sample sizes the uncertainty due to random sampling error usually is large and may become the dominant source of uncertainty in the output. This uncertainty could be reduced if there is relevant prior information, for example, reasonable estimates for distribution parameters from well-described datasets (Aldenberg and Luttik 2002). 

Conjugate pair In Bayesian estimation, when the observation of new data changes only the parameters of the prior distribution and not its statistical shape (i.e., whether it is normal, beta, etc.), the prior distribution on the estimated parameter and the distribution of the quantity (from which observations are drawn) are said to form a conjugate pair. In case the likelihood and prior form a conjugate pair, the computational burden of Bayes rule is greatly reduced. 

---