Using the same data as in Example A.5, estimate the density function using an exponential distribution. Test the hypothesis that this is a reasonable estimate.
The mean of the data is μ = 7. Since the mean is also the MLE for the
exponential distribution, the estimated density function is
\ \hat{f}(x)=\begin{cases} \frac{1}{7}e^{-x/7},&x≥0,\\0,&x<0. \end{cases}
Notice that for an exponential distribution, one must use an additional interval x ≥ 15 since the exponential is right-tailed. The expected frequencies are calculated as follows:
\ e_1=n\int_{0}^{2.5}{\hat{f}(x) dx}\\= 10\int_{0}^{2.5}{\frac{1}{7}e^{-x/7} dx}\\=-10e^{-2.5/7}|^{2.5}_0\\=10(1-e^{-2.5/7})\approx 3.00,\ e_2= 10\int_{2.5}^{5.0}{\frac{1}{7}e^{-x/7} dx}\approx 2.101,
\ e_3= 10\int_{5.0}^{7.5}{\frac{1}{7}e^{-x/7} dx}\approx 1.470,
\ e_4= 10\int_{7.5}^{10.0}{\frac{1}{7}e^{-x/7} dx}\approx 1.029,
\ e_5= 10\int_{10.0}^{12.5}{\frac{1}{7}e^{-x/7} dx}\approx 0.720,
\ e_6= 10\int_{12.5}^{15.0}{\frac{1}{7}e^{-x/7} dx}\approx 0.506,
\ e_7= 10\int_{15.0}^{\infty}{\frac{1}{7}e^{-x/7} dx}\approx 1.172.
Again calculating\ \chi^2,
\ \chi^2=\sum\limits_{k=1}^{7}{\frac{(\phi_k – e_k)^2}{e_k}}\\[0.3cm] = \frac{(3-3.000)^2}{3.000}+\frac{(2-2.101)^2}{2.101}+\frac{(1-1.470)^2}{1.470}+\frac{(0-1.029)^2}{1.029} \\+\frac{(2-0.720)^2}{0.720} +\frac{(2-0.506)^2}{0.506}+\frac{(0-1.172)^2}{1.172}\approx 9.043.Since there is but one parameter determined\ (l = 1) and 7 classes (m = 7), this time the number of degrees of freedom is\ \nu = 7 – 1 – 1 = 5 . Checking the critical values,\ \chi^2_{0.05} = 11.070 and \chi^2_{0.01} = 15.086 . Since\ \chi^2 < \chi^2_{0.05} and \chi^2 < \chi^2_{0.01} , we will accept this hypothesis at either the 5% or 1% significance level.