Fix/improve group_by, ungroup usage in most vignettes

* Add missing `group_by` before some `epix_slide`s that previously relied on a
  nontrivial default grouping, fixing vignette regression.
* Be more consistent about `ungroup`ing after grouped operations that don't
  automatically ungroup. Try not to introduce regressions in (uncommon) code that
  takes a grouped object from an earlier computation and performs manipulations
  that actually do rely on it being grouped.

Author: lcbrooks

vignettes/advanced.Rmd

@@ -62,12 +62,16 @@ df %>%
 df %>% 
   mutate(version = time_value) %>% 
   as_epi_archive() %>%
-  epix_slide(x_2dav = mean(x), before = 1, ref_time_values = as.Date("2020-06-02"))
+  group_by(geo_value) %>%
+  epix_slide(x_2dav = mean(x), before = 1, ref_time_values = as.Date("2020-06-02")) %>%
+  ungroup()
 
 df %>% 
   mutate(version = time_value) %>% 
   as_epi_archive() %>%
-  epix_slide(~ mean(.x$x), before = 1, ref_time_values = as.Date("2020-06-02"))
+  group_by(geo_value) %>%
+  epix_slide(~ mean(.x$x), before = 1, ref_time_values = as.Date("2020-06-02")) %>%
+  ungroup()
 ```
 
 When the slide computation returns an atomic vector (rather than a single value)
@@ -102,7 +106,8 @@ object returned by `epi_slide()` has a list column containing the slide values.
 df2 <- df %>% 
   group_by(geo_value) %>%
   epi_slide(a = data.frame(x_2dav = mean(x), x_2dma = mad(x)),
-            before = 1, as_list_col = TRUE)
+            before = 1, as_list_col = TRUE) %>%
+  ungroup()
 
 class(df2$a)
 length(df2$a)
@@ -120,7 +125,8 @@ slide computation (here `x_2dav` and `x_2dma`) separated by "_".
 df %>% 
   group_by(geo_value) %>%
   epi_slide(a = data.frame(x_2dav = mean(x), x_2dma = mad(x)),
-            before = 1, as_list_col = FALSE)
+            before = 1, as_list_col = FALSE) %>%
+  ungroup()
 ```
 
 We can use `names_sep = NULL` (which gets passed to `tidyr::unnest()`) to drop
@@ -130,7 +136,8 @@ the prefix associated with list column name, in naming the unnested columns.
 df %>% 
   group_by(geo_value) %>%
   epi_slide(a = data.frame(x_2dav = mean(x), x_2dma = mad(x)),
-            before = 1, as_list_col = FALSE, names_sep = NULL)
+            before = 1, as_list_col = FALSE, names_sep = NULL) %>%
+  ungroup()
 ```
 
 Furthermore, `epi_slide()` will recycle the single row data frame as needed in
@@ -152,9 +159,11 @@ df %>%
 df %>% 
   mutate(version = time_value) %>% 
   as_epi_archive() %>%
+  group_by(geo_value) %>%
   epix_slide(a = data.frame(x_2dav = mean(x), x_2dma = mad(x)),
              ref_time_values = as.Date("2020-06-02"),
-             before = 1, as_list_col = FALSE, names_sep = NULL)
+             before = 1, as_list_col = FALSE, names_sep = NULL) %>%
+  ungroup()
 ```
 
 ## Multi-row outputs

vignettes/aggregation.Rmd

@@ -190,6 +190,7 @@ xt %>%
   as_epi_df() %>%
   group_by(geo_value) %>%
   epi_slide(cases_7dav = mean(cases), before = 6) %>%
+  ungroup() %>%
   filter(geo_value == "Plymouth, MA",
          abs(time_value - as.Date("2021-07-01")) <= 3) %>%
   print(n = 7)
@@ -198,6 +199,7 @@ xt_filled %>%
   as_epi_df() %>%
   group_by(geo_value) %>%
   epi_slide(cases_7dav = mean(cases), before = 6) %>%
+  ungroup() %>%
   filter(geo_value == "Plymouth, MA",
          abs(time_value - as.Date("2021-07-01")) <= 3) %>%
   print(n = 7)

vignettes/outliers.Rmd

@@ -117,6 +117,7 @@ x <- x %>%
     x = time_value, y = cases,
     methods = detection_methods,
     combiner = "median")) %>%
+  ungroup() %>%
   unnest(outlier_info)
 
 head(x)