Exploring the effect of non-response in modeling the age at diagnosis in HPV-Related Cancers
Document typeMaster thesis
Rights accessRestricted access - author's decision
The motivation of this master's thesis emerges in the context of a worldwide cross‐ sectional study. The main goal of this work it is to study the effect of partial information data in the analysis of the age at diagnosis of the anogenital and head and neck cancers in presence of determinant genotypes of HPV. The ﬁrst aim is to present a methodology for dealing with the presence of partial information data. The second goal is to apply this methodology in a particular case of head and neck anatomical sub-locations. The methodology will be used for dealing with partial information in the target variable age at diagnosis. For this purpose, data will be multiple imputed under the assumed condition of Missing At Random (MAR). The methodology will also be used for dealing with partial information in a covariate, in this case the gender. For achieving this, it will be proposed a ﬁrst sensitivity analysis to explore departures from MAR non-response mechanism. The non-response model is governed by a one parameter of non-ignorability. The last aim will be to illustrate the eﬀect that has this partial information on the obtainment of the parameter estimates of interest (the average age at diagnosis), when assuming diﬀerent non-response mechanisms.