BI-CHI: A COMPUTER PROGRAM FOR THE CHI-SQUARE GOODNESS OF FIT TEST FOR A BIVARIATE NORMAL DISTRIBUTION

A description is given of the BI-CHI Bivariate Chi-Square Test computer program. The program is for assessing the agreement of a continuous two-dimensional sample distribution with a parent bivariate normal distribution. The program computes the Chi-square statistic for testing the null hypothesis The random sample is from a bivariate normal parent population. The minimum and maximum sample values of each random variable are determined, and the sample estimates of the five parameters two means, two variances, and the correlation coefficient of the bivariate normal distribution are computed. Also, the sample estimates of the two regression lines x sub two on x sub one and x sub one on x sub two are computed. An optional feature allows additional output of a plot of the sample data and the contour ellipses. In addition to an example problem illustrating the input and the output format, a flow chart and program listing are included in the report. The program is coded in FORTRAN IV for the IBM 7030 STRETCH computer.