الفهرس | Only 14 pages are availabe for public view |
Abstract One of the most important problems in frequency analysis is the estimation of the proper shape of the population for a given sample of data. The goodness of fit method is an effective means of examining how well a sample of data agrees with an assumed probability distribution as its population. The much more important situation is that in which F0(x) depends on unknown parameters. In this case the test can be modified by inserting estimates of parameters in F0(x), but their distribution theory is then much more complex. The distribution test statistics also depend on F0(x) so the tests are no longer distribution free. There is generally no satisfactory way to get percentage points for small n except by simulation though a few special available results. Percentage points have been obtained through Monte Carlo methods for the cases in which F0(x) is Exponential, Extreme values or Normal distributions and some other distributions. When data are Type-II censored, simple modification can be made to the empirical distribution function (EDF) goodness techniques of fit statistics and distribution theory becomes more complicated than in the corresponding uncensored situation. There are many goodness techniques for fit tests. Tables of critical values for this test are inappropriate when the parameters of the hypothesized distribution are unknown or when the sample is censored. Most goodness of fit statistics can be regarded as measures of proximity between two distributions: empirical and hypothesized |