![Sample our Humanities journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5939/TOC-Subject-Sample-2021-Humanities.jpg"})
Notes
This paragraph summarizes a more detailed discussion of consistency with prior work in Blank (Citation2013, pp. 603–606). This note is the place to deal with a misunderstanding. It is true that Hargittai and Walejko's (Citation2008) Table 7 contains only respondents who create content, but Ms Schradie overlooks their footnote 5, which says that ‘results are robust when the analyses are performed on the entire sample’ (p. 253). Logistic regression analyses on the full sample would contrast respondents who post online content versus those who do not, which is exactly what my Table 3 does (Blank Citation2013). My Note 7 (Blank Citation2013) makes this point, but Note 7 does not clarify that when it refers to Table 3 it means the table in my article and not the Table 3 in Hargittai and Walejko's (Citation2008) article.
Ms Schradie says that I misunderstood our email exchange in Note 9 (p. 610). In my email I asked why, when Table 3 was based on Model 6, the significance levels did not match the significance levels for Model 6 in the online appendix (Schradie Citation2011b, see Tables G–P). Readers can easily confirm that there is a big difference. Table 3 reports six significant coefficients for ‘H.S. grad’, whereas Model 6 in the appendix shows only 2 of 10 are significant. She replied that the significance levels are from Model 5, while the predicted probabilities reported in Table 3 are from Model 6. I leave it to readers to decide whether it is a valid use of statistics to report significance levels from one model attached to parameter estimates based on a different model.
Only two possible variables including year (‘Time’ and ‘Time × H.S. Grad’) appear in Tables G–P in Schradie (Citation2011b). Ms Schradie's description of her methodology is minimal and she never describes these variables, so we do not know what they are. However, both are reported as single coefficients, not individual coefficients for each year, which they would be if they were dummy variables for each year and interaction effects by each year.
This paragraph summarizes a more detailed discussion on pages 605–606 in my paper (Blank Citation2013). See especially Figure 2 and the accompanying text for an explanation of why the difference between within-survey and between-survey coefficients is a critical issue.
My paper at least has three variables that Ms Schradie concedes measure actual content: the two political variables, and the variable measuring ‘writing and other creative content’.
Actually, the online content creation questions will never be asked of offline respondents. In the Pew surveys they have missing values for those questions. Ms Schradie presumably changed the missing values to zeros, although she does not say this in her article.
An exception is my comments on Ms Schradie's combination of multiple surveys into a single data-set, which is a critique.
