home *** CD-ROM | disk | FTP | other *** search
-
- _C_o_m_p_u_t_e _R_e_s_i_d_u_a_l_s _f_o_r _s_u_r_v_r_e_g _O_b_j_e_c_t_s
-
- residuals.survreg(object, type)
- This is a method for the function residuals() for objects inheriting from
- class glm. Several types of residuals are available for glm objects,
- hence the additional argument:
-
- _A_r_g_u_m_e_n_t_s:
-
- type:
- type of residuals, with choices "deviance", "pearson",
- "working" or `"matrix"; the first is the default.
-
- Value:
-
- A vector of residuals is returned. The sum of squared
- deviance residuals add up to the deviance. The pearson
- residuals are standardized residuals on the scale of
- the response. The working residuals reside on the
- object, and are the residuals from the final IRLS fit.
- The matrix type produces a matrix based on derivatives
- of the log-likelihood function. Let L be the log-
- likelihood, p be the linear preditor X %*% coef, and s
- be log(sigma). Then the 6 columns of the matrix are L,
- dL/dp, ddL/(dp dp), dL/ds, ddL/(ds ds) and ddL/(dp ds),
- where d stands for the derivative and dd the second
- derivative. Diagnstics based on these quantities are
- dicussed in an article by Escobar and Meeker.
-
- References:
-
- Escobar and Meeker (1992). Assessing influence in
- regression analysis with censored data.
- (
- c
- k
- s
- l
- sh)fIBiometrics,(
- c
- k
- s
- l
- sh)fP 48, 507-528.
-
- _E_x_a_m_p_l_e_s:
-
- fit <- survreg(Surv(time,status) ~x, aml)
- rr <- residuals(fit, type='matrix')
-
-