Corrigendum: A Multifaceted Sensitivity Analysis of the Slovenian Public Opinion Survey Data

Abstract

There are some errors in the text which affect the interpretation, as follows. The penultimate paragraph on page 172 is incorrect and should read We focus on two binary questions, such as the independence and attendance questions in the SPO survey. The 16 theoretical full cell probabilities are as in Table 2, part (a), producing 15 full data degrees of freedom. The generic expression for the cell probabilities is πr1r2,j1j2=P(R1=r1, R2=r2, J1=j1, J2=j2), where R1 and R2 denote the missingness indicators for the attendance and independence questions respectively, whereas J1 and J2 denote the actual responses for the attendance and independence questions respectively. Further, r1=0 (or r1=1) if the answer to the attendance question is missing (or observed), and j1=1 (or j1=2) if the answer to the attendance question is yes (or no). The indices r2 and j2 are similarly defined for the independence question. In the last paragraph on page 175, the words ‘independence’ and ‘attendance’ should be interchanged on both occasions so that the sentence reads The α- and β-parameters describe missingness respectively in the attendance and independence questions as the proportion of subjects with a missing response on the attendance and independence questions relative to the proportion of subjects with both responses present, given a particular response combination (j1,j2). On page 176, for consistency of notation the second column in Table 4 should contain the following entries instead: In the last sentence of page 182, cell 5 has been incorrectly described. The sentence should read Similarly, exclusion of a single respondent with a yes on the attendance question but a missing response on the independence question (cell 5) also influences BRD4’s model parameters, though to a lesser extent. The last two sentences of Section 4.3 on page 183 are incorrect and should be Thus, adding a subject with a no on attendance and a missing independence response, or excluding a respondent with no on both questions, yields changes in the model parameters of BRD7 and BRD8. These findings hint at the influential nature of subjects with a no on the attendance question, which is likely to be related to this group's sparseness. On page 184, in the first line of text, the upper bounds p and q need to be replaced with s and N respectively, so that the phrase reads ‘with i,l=1,…,p and j=1,…,q’. In the last equation on page 185 a subscript is missing on the left-hand side, i.e. ‘qr1r2|j2’ should be ‘qr1r2|j12’. The reference to the table in the caption to Fig. 5 on page 186 should be to Table 2, part (a), and not to Table 1 as stated. Lastly, the sentence starting at the bottom of page 191 is incorrect and should read Note that small perturbations in cell 3 seem to affect only the predicted cell counts in the top row (j1=1; yes on attendance) under model BRD5 and/or BRD6, whereas such changes impact the cell position (j1,j2)=(1,2) (yes on attendance—no on independence) under models BRD1, BRD2, BRD3, BRD6 and/or BRD9.