# Question: A statistician has to decide on the basis of one

A statistician has to decide on the basis of one observation whether the parameter θ of a Bernoulli distribution is 0, 1/2 , or 1; her loss in dollars (a penalty that is deducted from her fee) is 100 times the absolute value of her error.

(a) Construct a table showing the nine possible values of the loss function.

(b) List the nine possible decision functions and construct a table showing all the values of the corresponding risk function.

(c) Show that five of the decision functions are not admissible and that, according to the minimax criterion, the remaining decision functions are all equally good.

(d) Which decision function is best, according to the Bayes criterion, if the three possible values of the parameter θ are regarded as equally likely?

(a) Construct a table showing the nine possible values of the loss function.

(b) List the nine possible decision functions and construct a table showing all the values of the corresponding risk function.

(c) Show that five of the decision functions are not admissible and that, according to the minimax criterion, the remaining decision functions are all equally good.

(d) Which decision function is best, according to the Bayes criterion, if the three possible values of the parameter θ are regarded as equally likely?

## Answer to relevant Questions

A statistician has to decide on the basis of two observations whether the parameter θ of a binomial distribution is 1/4 or 1/2 ; his loss (a penalty that is deducted from his fee) is $ 160 if he is wrong. (a) Construct a ...With reference to Example 9.7, show that if the losses are proportional to the squared errors instead of their absolute values, the risk function becomes And its minimum is at k = 3/2. Example 9.7 A random variable has the ...If X1, X2, . . . , Xn constitute a random sample from a normal population with µ = 0, show that Is an unbiased estimator of σ2. The information about θ in a random sample of size n is also given by Where f (x) is the value of the population density at x, provided that the extremes of the region for which f(x) ≠ 0 do not depend on θ. The ...With reference to Example 10.3, we showed on page 281 that – 1 is an unbiased estimator of d, and in Exercise 10.8 the reader was asked to find another unbiased estimator of d based on the smallest sample value. Find the ...Post your question