When Z is a continuous variable, such as income, summation is replaced by integration and the probabilities are replaced by the appropriate density function. Where Z m are the Q values that the variable Z can take, μ is the expected value of Z, and Prob ( Z m) are the associated probabilities. We end by considering the special case of sample averages of functions of interest.
#CALCULATE STANDARD ERROR OF ESTIMATE SOFTWARE#
We discuss the advantages and disadvantages of each method and provide sample computer code for two popular software packages in the Appendix. In our experience, the three methods that we have chosen to discuss are the most common in the health services research literature and can be applied regardless of the method used to estimate the coefficients in the function of interest. Some methods are integral to the estimation of the coefficients that subsequently appear in the function of interest, such as the method of moments and Gibbs sampling. There are other ways to compute standard errors. Krinsky and Robb or K–R (Krinsky and Robb 1986, 1990) We discuss three methods of computing the standard errors of functions of interest: We then discuss standard errors in the context of a simple linear model, before turning to more complex nonlinear models. This is a central issue in the question raised in the previous paragraph. The article begins with a brief note on the distinction between a standard deviation and a standard error. For example, is the variation of interest the variance that arises from inserting different values of the explanatory variables into the function the fact that coefficients in the function are estimated parameters or both? The construction of confidence intervals requires estimation of the variance of the function, which in turn requires careful consideration of the sources of variation in the function's value. Confidence intervals allow the analyst to test hypotheses about the value of the function-for example, to evaluate the proportion of the function's values that would fall within a given range, if the “experiment” were repeated multiple times. While the computation of the function itself often is straightforward, establishing confidence intervals for the function's value can be more difficult. Common examples of functions of estimated parameters include the predicted value of the dependent variable for a particular subject or set of subjects in the data, and the effect of a change in an explanatory variable on the predicted value of the dependent variable (sometimes referred to as a partial effect, marginal effect, or incremental effect) and elasticities. The questions posed by standard analyses in health services research often require evaluation not only of estimated parameters (e.g., regression coefficients) but also functions of estimated parameters.
0 kommentar(er)
