Frequency Analysis

Counts: statlist = statlist(“n”)

There are many uses for freq, the first one we’ll show is to create the first row of analysis set summary of population counts in the output. In the following table the subset argument is used to create the one row summary with the row heading specified in the rowtext argument.

tbl <- cdisc_adsl %>% 
  freq(colvar = "TRT01PN", 
       rowvar = "ITTFL", 
       statlist = statlist("n"),
       subset = ITTFL == "Y",
       rowtext = "Analysis set: ITT")
  
knitr::kable(tbl)

Big N row, counts and percentages: statlist = statlist(c(“N”, “n (x.x%)”))

A typical call of using the freq function is to create the tbl chunk with:

• the row of big N indicating number of subjects having values in the character variable for each treatment group

• summary of counts and percentages for each category in the character variable such as age groups, gender, race, ethnicity, etc.

tbl <- cdisc_adsl %>% 
  freq(colvar = "TRT01PN", 
       rowvar = "SEX", 
       statlist = statlist(c("N","n (x.x%)")),
       row_header = "Sex")

knitr::kable(tbl)

Using factor to order category summary

In the above table, F appears first and then M is shown. This is because SEX is a character variable and alphabetical sorting is applied to the summarized results. We can convert SEX to a factor variable with the customized order and labels for enabling the user-defined sorting.

tbl <- cdisc_adsl %>% 
  mutate(SEX = factor(SEX, levels = c("M", "F"), labels = c("Male", "Female"))) %>% 
  freq(colvar = "TRT01PN", 
       rowvar = "SEX", 
       statlist = statlist(c("N","n (x.x%)")),
       row_header = "Sex")

knitr::kable(tbl)

Counts, denominators, and percentages: statlist = statlist(c(“n/N (x.x%)”))

In some outputs, the table mock-up may require showing the percentage denominator, which can be done by specifying statlist = statlist("n/N (x.x%)").

tbl <- cdisc_adsl %>% 
  mutate(SEX = factor(SEX, levels = c("M", "F"), labels = c("Male", "Female"))) %>% 
  freq(colvar = "TRT01PN", 
       rowvar = "SEX", 
       statlist = statlist("n/N (x.x%)"),
       row_header = "Sex")

knitr::kable(tbl)

By processing: statlist = statlist(c(“N”, “n (x.x%)”))

By-processing splits the summary statistics by another character variable that can be specified in the argument of rowbyvar or tablebyvar. For the example shown below, age group categories are summarized by the SEX variable. In this scenario, the denominator should also be split by SEX in addition to TRT01PN, which can be done by statlist(c("N","n (x.x%)"), denoms_by = c("SEX", "TRT01PN")).

tbl <- cdisc_adsl %>% 
  mutate(SEX = factor(SEX, levels = c("M", "F"), labels = c("Male", "Female"))) %>% 
  freq(colvar = "TRT01PN",
       rowbyvar = "SEX",
       rowvar = "AGEGR1", 
       statlist = statlist(c("N","n (x.x%)"), denoms_by = c("SEX", "TRT01PN")),
       row_header = "Age group") 

knitr::kable(tbl)

By default, the denominators are calculated by using colvar, tablebyvar, and rowbyvar. The above freq function call will also produce the same results without specifying the denoms_by argument inside the statlist function.

Padding factor levels on and off

When using rowbyvar to create by-processing summaries, some levels of the rowvar may have zero records as shown in the example below.

tbl <- cdisc_adsl %>% 
  freq(colvar = "TRT01PN",
       rowbyvar = "ETHNIC",
       rowvar = "RACE",
       statlist = statlist(c("N","n (x.x%)")),
       row_header = "Race")

knitr::kable(tbl)

To remove the zero record rows and create the data driven summary, users can specify the pad = FALSE in the freq function.

tbl <- cdisc_adsl %>% 
  freq(colvar = "TRT01PN",
       rowbyvar = "ETHNIC",
       rowvar = "RACE",
       statlist = statlist(c("N","n (x.x%)")),
       row_header = "Race",
       pad = FALSE)

knitr::kable(tbl)

Denominator Dataframe

When using frequency analysis you do not need to always use your main dataframe to calculate your denominators. For example, an adverse event table may use ADSL as the denominator dataframe instead of ADAE even though the counts are coming from ADAE. The example below shows a table counting AEDECOD but using ADSL to calculate the denominators.

adae <- cdisc_adae %>%
  rename(TRT01AN = TRTAN)

tbl <- adae %>%
  freq(denom_df = cdisc_adsl,
       colvar = "TRT01AN",
       rowvar = "AEDECOD",
       descending_by = "81") 

knitr::kable(head(tbl, 10))

By processing and denominators: statlist = statlist(c(“n (x.x)”), denoms_by = “TRT01AN”)

When using by variables such as tablebyvar and rowbyvar the denominators may need to be changed to work correctly. This also works in conjunction with the denom_df argument. By default, the denominators are calculated by using colvar, tablebyvar, and rowbyvar. This works well if you are using the same dataframe for counts and denominators but this is not always the case. In the following example we are doing a similar table as above but using AESEV as a rowbyvar. In this example we don’t want our denoms by AESEV since that variable is not in ADSL, which is where our denominators are coming from. To change the variables by which our denoms are calculated by we use the denoms_by argument to the statlist function. Below you can see that we are using only our colvar of TRT01PN as our denoms_by.

tbl <- adae %>%
  freq(denom_df = cdisc_adsl,
       colvar = "TRT01AN",
       rowvar = "AEDECOD",
       rowbyvar = "AESEV",
       statlist = statlist(c("n (x.x)"), denoms_by = "TRT01AN")) 

knitr::kable(head(tbl, 10))

Available statlist formats for frequency summary

We have shown the common use cases of calling different variants of the statlist for frequency summaries. The table below describes all available options to be specified for the statlist in freq.

To learn more about using the statlist function for freq analysis, please type ?statlist in your console.