If the two regressors are strongly associated, the data may provide little support for varying one while holding the other fixed. Multicollinearity makes the separate slope estimates unstable, and exact linear dependence makes them unidentifiable. A conditional slope may also be an extrapolation beyond combinations actually observed.
Furthermore, a slope in a linear regression model is not automatically a causal effect. A causal interpretation needs appropriate design or assumptions excluding confounding; independence of the errors across observations and zero mean alone do not establish those assumptions. The coefficients describe the specified linear mean relation, but changing a regressor by intervention can have a different interpretation.
The centered regressors make the first normal equation read . DefineThe remaining equations are and . Provided , matrix inversion givesBy the Cauchy-Schwarz inequality, . If , the two regressor columns are linearly dependent, and their separate slopes are not uniquely identifiable. One still has well-defined linear least squares fitted values, but cannot use the displayed divisions by .
Minimize the residual sum of squaresSetting its three partial derivatives to zero gives the normal equationsThese equations characterize the linear least squares minimizers, since is a convex quadratic. A unique coefficient vector additionally requires full column rank of the design matrix.
Since , we have . Squaring and adding the other residual squares givesTherefore is a measurable function of . By the block independence proved in part i, it is independent of . This is the independence of the normal sample mean and sample variance; pairwise independence alone would not justify passing to an arbitrary function of several residuals.
Each residual has mean zero and variance . By linearity of expectation, without needing independence of residuals,Hence is an unbiased estimator of .
Put . The supplied linear-transformation result shows that is jointly multivariate normal. Its marginal distributions areIndeed, , soFor a jointly Gaussian vector, zero cross-covariance between blocks implies their independence. Thus is independent of the entire vector , not merely of each coordinate separately. The residual coordinates themselves need not be independent: for , their covariance is .
The sample mean is exactly , so has the standard normal distribution. The given quantile gives probability to the interval . Inverting this pivot yields the exact confidence interval
For known , maximizing the normal likelihood over is equivalent to minimizing . The identityshows that the unique minimum occurs at . Equivalently, the log-likelihood has second derivative .
The central limit theorem givesWrite for the indicated standard normal quantile. Inverting the approximate pivot inequality and substituting gives the approximate confidence intervalIts coverage is approximately for large . The intersection respects the positive rate parameter; asymptotically the lower endpoint is positive. This is an asymptotic interval, not an exact finite-sample confidence interval.
The inverse moments in part i giveThus the bias is positive, and direct expansion of the mean squared error yieldsFor comparison, the variance is ; adding the squared bias gives the same answer by the bias-variance decomposition.
Use the gamma density and the normalization integral for shape :The same argument, valid for , gives .
For the independent random sample, the likelihood function on its support is , where . Its log-likelihood is , withAlmost surely , and the unique stationary point is consequently the unique global maximum:The exceptional all-zero sample has probability zero; if were nevertheless supplied, there would be no finite maximizing rate.
For an estimator of , its bias and mean squared error areWhen the second moment exists, expanding around gives the bias-variance decomposition .
For the exponential distribution with rate , the change of variables givesThe integrals are one and two by integration by parts. Hence .
By part iii, is an integer with probability zero. Thus almost surely, andThe pointwise inequalities implyPutting and rearranging yieldsThis extracts bounds for an infinite series from simple rounding inequalities.
For a nonnegative integer-valued random variable , the tail-sum formula states . It follows by writing and using nonnegativity to interchange sum and expectation.
For an integer , is the same event as . Integration for the Pareto distribution givesThe second identity subtracts two successive tail probabilities, equivalently integrating over . If includes zero, the extra case is and , since almost surely.
Normalization of the probability density function requiresHence . The distribution is a Pareto distribution with minimum one and shape parameter .
The integers form a countable set, and the events are disjoint. By countable additivity and part ii,Countability is essential: one cannot use the same sum argument over an arbitrary uncountable collection of zero-probability events.
There are unlisted articles, also show them or only show them.