![]() Of course p-values can be low for reasons other than a too low k. Currently smooths of factor variables are not supported and will give an NA p-value.ĭoubling a suspect k and re-fitting is sensible: if the reported edf increases substantially then you may have been missing something in the first fit. Note the disconcerting fact that if the test statistic itself is based on random resampling and the null is true, then the associated p-values will of course vary widely from one replicate to the next. Low p-values may indicate that the basis dimension, k, has been set too low, especially if the reported edf is close to k', the maximum possible EDF for the term. ![]() For models fitted to more than k.sample data, the tests are based of k.sample randomly sampled data. The p-value is computed by simulation: the residuals are randomly re-shuffled k.rep times to obtain the null distribution of the differencing variance estimator, if there is no pattern in the residuals. The further below 1 this is, the more likely it is that there is missed pattern left in the residuals. This estimate divided by the residual variance is the k-index reported. The test of whether the basis dimension for a smooth is adequate (Wood, 2017, section 5.9) is based on computing an estimate of the residual varianceīased on differencing residuals that are near neighbours according to the (numeric) covariates of the smooth. For the default optimization methods the convergence information is summarized in a readable way, but for other optimization methods, whatever is returned by way ofĬonvergence diagnostics is simply printed. Usually the 4 plots are various residual plots. This function plots 4 standard diagnostic plots, some smoothing parameter estimationĬonvergence information and the results of tests which may indicate if the smoothing basis dimension For these reasons this routine uses an enhanced residual QQ plot. Similarly, it is not clear how sensitive REML and ML smoothness selection will be to deviations from the assumed response dsistribution. For example, un-modelled overdispersion will typically lead to overfit, as the smoothness selection criterion tries to reduce the scale parameter to the one specified. ![]() In particular, the thoery of quasi-likelihood implies that if the mean variance relationship is OK for a GLM, then other departures from the assumed distribution are not problematic: GAMs can sometimes be more sensitive. Secondly, fitting may not always be as robust to violation of the distributional assumptions as would be the case for a regular GLM, so slightly more care may be needed here. choose.k provides more detail, but the diagnostic tests described below and reported by this function may also help. Oversmoothing: the defaults are arbitrary. The basis dimensions used for smooth terms need to be checked, to ensure that they are not so small that they force How many re-shuffles to do to get p-value for k testing.Īrguments passed to qq.gam() when old.style isĮxtra graphics parameters to pass to plotting functions.Ĭhecking a fitted gam is like checking a fitted glm, with two main differences. ![]() Type of residuals, see residuals.gam, used inĪbove this k testing uses a random sub-sample of data. If you want old fashioned plots, exactly as in Wood, 2006, set to TRUE. )Ī fitted gam object as produced by gam(). Care should be taken in interpreting the results when applied to gam objects returned by gamm. Plots, some information about the convergence of the smoothness selection optimization, and to runĭiagnostic tests of whether the basis dimension choises are adequate. Takes a fitted gam object produced by gam() and produces some diagnostic informationĪbout the fitting procedure and results. Some diagnostics for a fitted gam model Description ![]()
0 Comments
Leave a Reply. |