2020-2022 Social Media Study: Updates & Errata
Current Dataset Release: July 5, 2023 version
The study documentation for the November 18, 2021 version of the dataset describes the post-election weights as having been trimmed to a maximum value of 5. However, the released dataset’s weights were not trimmed, and 40 cases have weight values exceeding 5.0.
The code for these data changes is:
If weight_post > 5 weight_post = 5.
Compute weight_post = weight_post /0.989108 .
Compute weight_post_spss = weight_post * .61 .
The July 5, 2023 version of the dataset has been updated to include this change.
The post-election questionnaire shows incorrect logic to select
participants to receive the questions w2nd_fbd and w2nd_fbr. The correct
logic as implemented in the study was “[IF (w2fbuser=1 OR 2) or
(fbuser=1 OR 2)]”.
Items w2discern, w2fb1, and w2serious were programmed to be asked of all
respondents but 98, 102, and 131 participants, respectively, are coded
-1, Inapplicable. These should be indicated as -5, “Interview breakoff
(sufficient partial IW)” instead of inapplicable.
Items w2fb2 through w2fb5 have 105 to 125 breakoff cases incorrectly
coded as -1, inapplicable, among a larger number of correctly coded
inapplicable cases. Breakoff cases for these four variables can be
recoded to -5 with the following Stata code:
replace user =1 if fbuser==1 | fbuser==2 | w2fbuser==1 | w2fbuser==2
replace w2fb2 = -5 if w2fb2== -1 & user==1
replace w2fb3 = -5 if w2fb3== -1 & user==1
replace w2fb4 = -5 if w2fb4== -1 & user==1
replace w2fb5 = -5 if w2fb5== -1 & user==1
In the Post-election questionnaire, four corrections are needed.
Items w2fb4 and w2fb5 indicated response option 5 as “All the time,” but the actual response option displayed to participants was “Always”.
Item w2covid_know indicated the response options were “I know someone with the disease” and “I do not know anyone with it.” The correct response options, respectively, are “I know someone who has had the disease” and “I do not know anyone who has had it”.
Item w2covid_number indicated response option 4 as “Slightly fewer people have died” but the correct response wording was “Somewhat fewer people have died”.