On content analysis of images of mass protests: a case of data triangulation

The development of automated face recognition technologies raises concerns as to the protection of the privacy of individuals whose pictures are freely available on the internet (Risen and Poitras 2014).

The agreement criterion used is the frequency of codes in case, which complicates the task of reaching an acceptable level of agreement compared with a less stringent—yet more commonly reported one (the presence or absence of codes in case).

The code book had initially contained separate codes corresponding to five age groups: 20 years old and younger, 21–30 years old, 31-40 years old, 41–50 years old and 51 plus years old; they were subsequently transformed into four age groups: 24 years old and younger, 25–39 years old, 40–54 years old, 55 plus years old.

A possible explanation may refer to the unwillingness to disclose personal information, including demographic data, on-line that characterizes collectivistic cultures (Cho and Park 2013, pp. 2325–2326). Russian culture has several collectivistic features. The other explanation refers to a low level of the validity of personal information posted on-line in general.

The respondent could select several options when answering a survey question about his/her ideological affinities, hence the total count in the “Survey” column exceeds 100 %.

A binary logistic regression analysis shows that the supporters of nationalist ideas tend to be relatively younger and less educated than the other protesters. The outcome variable is a stated affinity with nationalist ideas (0 = No stated affinity, 1 = Stated affinity). Six predictor variables were included in the model. The first refers to the respondent’s gender (0 for females, 1 for males). The second refers to the respondent’s age at the scale level of measurement. The third refers the respondent’s education (0 for people with no university education, 1 for those with some university education). The fourth is a response to the survey question about whether or not the respondent learned about the December 24 demonstration from the internet (0 for those who did not and 1 for those who did). The fifth refers to the respondent’s relative well-being (0 for people who cannot afford to buy a car, 1 for those who can). The sixth refers to the respondent’s liberal values (0 for people who do not identify themselves as liberals, 1 for those who do). The binary regression logistic regression procedure in SPSS was used to perform the analysis. The method for entering predictors was Forward Stepwise (Likelihood Ratio). In Step 1, Age was entered. It produced a Cox and Snell’s \(R^{2}\) increment of .02 and a Nagelkerke’s \(R^{2}\) increment of 0.055. In Step 2, Education was entered. It produced a Cox and Snell’s \(R^{2}\) increment of 0.01 and a Nagelkerke’s \(R^{2}\) increment of 0.025. A test of the full model with the two predictor variables, Age and Education, compared with a constant-only or null model was statistically significant, \(\chi ^{2}(2)=24.105,\, p<.001\). The strength of the association between, on the one hand, the affinity with nationalist ideas and, on the other hand, Age and Education, was not strong with Cox and Snell’s \(R^{2}=0.03\) and Nagelkerke’s \(R^{2}=0.082\). The sign of B for Age is negative, that for Education is positive. In other words, the relative prevalence of older protesters suggested by the survey could have resulted from the relative underrepresentation of young protesters with nationalist ideas in its sample.

They were expressed by one of the reviewers.