USSC survey analysis

The intended purpose for this write-up is to explain the two new weighted prop.table functions in ussc.

The first, ussc::prop_survey_question(), calculates the weighted proportion for specified questions across both countries and spreads the results so that the resulting data frame is wide (i.e. corresponding US and Australian proportions for the question and answer are side-by-side). It works as follows:

f1 <- survey_data %>%
  prop_survey_question(questions = starts_with("stance_on_tarrifs")) %>%
  select(-description_au) %>%
  kable("html", digits = 0, booktabs = TRUE)

f1

description_us	answer	Australia	United States	Difference
stance on tariffs (general)	Bad for the American economy	32	42	10
stance on tariffs (general)	Good for the American economy	25	27	2
stance on tariffs (general)	Neither good nor bad for the American economy	43	31	-12
stance on tariffs (trump influence)	Bad for the American economy	40	41	1
stance on tariffs (trump influence)	Good for the American economy	24	33	9
stance on tariffs (trump influence)	Neither good nor bad for the American economy	36	26	-10

The sample group has already been specified for you. You do need to make note of the particular questions you want to select.

Keep in mind that the answer column has been coerced to a character variable. To turn it back into a factor, run the following function relevel_survey_answer:

survey_data %>%
  prop_survey_question(questions = starts_with("stance_on_tarrifs")) %>%
  relevel_survey_answer(c("Good for the American economy",
                          "Neither good nor bad for the American economy",
                          "Bad for the American economy")) %>%
  select(-description_au) %>%
  kable("html", digits = 0, booktabs = TRUE)

description_us	answer	Australia	United States	Difference
stance on tariffs (general)	Good for the American economy	25	27	2
stance on tariffs (general)	Neither good nor bad for the American economy	43	31	-12
stance on tariffs (general)	Bad for the American economy	32	42	10
stance on tariffs (trump influence)	Good for the American economy	24	33	9
stance on tariffs (trump influence)	Neither good nor bad for the American economy	36	26	-10
stance on tariffs (trump influence)	Bad for the American economy	40	41	1

Your data frame will reorder based on the factor levels in the answer column.

The second function, ussc::prop_grouped_survey_question() allows the user to group by two or more variables, like so:

f2 <- survey_data %>%
  prop_grouped_survey_question(questions = starts_with("sexual_harrassment"), partyid) %>%
  select(-description_au) %>%
  rename_all(~sub('Australia_|(United States_)', '', .x)) %>%
  rename("AU other/none" = `Other or none`)

knitr::kable(f2, digits = 0, booktabs = TRUE)

description_us	answer	Coalition	Greens	Labor	AU other/none	Democrat	Independent	Republican
degrees of sexual harrassment - men getting away with committing sexual harassment and assault	A major problem	62	76	70	59	83	60	45
degrees of sexual harrassment - men getting away with committing sexual harassment and assault	A minor problem	32	22	24	32	14	30	44
degrees of sexual harrassment - men getting away with committing sexual harassment and assault	Not a problem	6	1	5	9	2	11	12
degrees of sexual harrassment - women making false claims about sexual harassment or assault	A major problem	47	26	37	46	32	46	58
degrees of sexual harrassment - women making false claims about sexual harassment or assault	A minor problem	46	42	50	46	43	41	36
degrees of sexual harrassment - women making false claims about sexual harassment or assault	Not a problem	8	32	13	8	25	14	6
degrees of sexual harrassment - women not being believed about sexual harassment and assault	A major problem	52	71	59	50	76	49	30
degrees of sexual harrassment - women not being believed about sexual harassment and assault	A minor problem	37	25	33	33	21	32	48
degrees of sexual harrassment - women not being believed about sexual harassment and assault	Not a problem	10	4	8	17	3	18	23

You can group by multiple variables, but keep in mind that the tables will get very wide and we will need to fit our output on a standard A4 paper.

Both functions return the description for the Australian survey and the American survey. The questions were often the exact same, so you can simply filter out a description column.

The column names will return information about the country and grouping variables. These, of course, should be cleaned up. I leave this up to the user.

Of course, you can pass other functions to the returned object. A pretty kable table can be pulled together quite quickly, e.g.

t <- f2 %>%
  select(-description_us,
         -`AU other/none`) %>%
  rename(" " = "answer") %>%
  knitr::kable("html", booktabs = T, digits = 0) %>%
  kableExtra::add_header_above(c(" " = 1,
                     "Australia" = 3,
                     "United States" = 3), bold = T) %>%
  kableExtra::group_rows(index = c("Men getting away with committing sexual harassment and assault" = 3,
                       "Women making false claims about sexual harassment or assault" = 3,
                       "Women not being believed about sexual harassment and assault" = 3))
t

	Australia			United States
	Coalition	Greens	Labor	Democrat	Independent	Republican
Men getting away with committing sexual harassment and assault
A major problem	62	76	70	83	60	45
A minor problem	32	22	24	14	30	44
Not a problem	6	1	5	2	11	12
Women making false claims about sexual harassment or assault
A major problem	47	26	37	32	46	58
A minor problem	46	42	50	43	41	36
Not a problem	8	32	13	25	14	6
Women not being believed about sexual harassment and assault
A major problem	52	71	59	76	49	30
A minor problem	37	25	33	21	32	48
Not a problem	10	4	8	3	18	23

Zoe Meers

2020-08-27