RISR, no. 21/2019
231 INTELLIGENCE, SECURITY AND INTERDISCIPLINARITY
the forward Wald selection test it was then tested in three steps which characteristics in this model were important (Lammers, 2007). In other words, the independent characteristics were added one by one, and for each it was tested whether the addition improved the model. This explains why the third column in Table 5 is relatively sparsely populated, which is due to the different selection procedure used for this test and the fact that the smaller data set used was the more accurate one. As a result this column contains only those characteristics for which tests have shown that they are significant predictors. The characteristics ‘hatred-revenge’ and ‘confusion’ in particular are significantly associated with the chance of being arrested on suspicion of a criminal offence. In addition, the characteristic of ‘fixation’ turned out to be less associated with those that were later arrested on suspicion of a criminal offence (Exp (B) <1). In the other analyses (Table 2, 3) fixation appeared mostly in connection with non-criminal threatening letters and with repeated letters. Furthermore the characteristic of ‘confusion’ turned out to be an important predictor for whether a letter writer would resort to writing more than one letter. Table 5 Regression analyses: threatening letters, repeated letters and arrest for criminal offences Threatening letter (n=278) Background characteristics Cognitive distortion Modi operandi Incoherent language Fixation Confusion Hatred-revenge
Exp (B) ,804
Sig. ,594
Repeated letter (n=133) Exp Sig. (B) 1,127 ,849
22,139 ,000** 1,515 ,376
1,125 ,870 ,584 ,360
,520 ,399 5,521
1,190 ,823 3,177 ,034* 1,400 ,622
,236 ,014* ,001**
Arrest for criminal offence (n=53) Exp Sig. (B)
,075 ,025* 13,529 ,005** 20,038 ,032*