02-first_steps.Rmd

# First Steps in R

In this chapter, you can find the basics of data wrangling in R for transforming, cleaning and exploring your data. 

Please take a look at this wonderful SAS2R **cheat sheet** created by **Brendan O’Dowd**. 

This one and many more cheat sheets can be found on the [RStudio website](https://www.rstudio.com/resources/cheatsheets/):

```{r out.height = "460px", out.width='800px', echo=F}
knitr::include_graphics("./additional resources/sas-r_cheatsheet.pdf")
```

## R and SAS Syntax

Below are some examples that use `adsl` to show how common operations are done in SAS and R.

### Packages and Sample Data
```{r, warning = F, message = F}
# Data
adam_path <- "https://github.com/phuse-org/TestDataFactory/raw/main/Updated/TDF_ADaM/"
adsl <- haven::read_xpt(paste0(adam_path, "adsl.xpt"))

# Select a few variables
adsl <- adsl %>%
  dplyr::select(STUDYID, USUBJID, SUBJID, AGE, TRT01P, TRTSDT, TRTEDT, RACE, SEX, DISCONFL)
```

### PROC CONTENTS $\rightarrow$ summary()
To explore the variables of your dataset you can use PROC contents in SAS: 

```{sas, eval = F}
proc contents data = adsl;
run;
```

In R, you can use `str()` and `summary()` instead:
```{r}
str(adsl)
summary(adsl) 
```


### PROC FREQ $\rightarrow$ count()
In order to get the frequencies for one variables you use PROC FREQ in SAS. 
```{sas, eval = F}
PROC FREQ data = adsl;
  TABLES SEX;
RUN;
```

And for cross tables: 
```{sas, eval = F}
PROC FREQ data = adsl;
  TABLES SEX * TRT01P * RACE;
RUN;
```

In R we use `table()` or `count()` from the tidyverse package:

**For One Variable**
```{r, message = F, eval = F}
table(adsl$SEX)

adsl %>%
  count(SEX)
```

```{r, message = F, eval = T, echo = F, warning = F}
library(kableExtra)
library(dplyr)

table(adsl$SEX)

adsl %>%
  count(SEX) %>%
  kbl() %>%
  kable_classic_2(full_width = F)
```


**For Multiple Variables**
```{r, message = F, eval = F}
#for multiple variables
table(adsl$SEX, adsl$TRT01P, adsl$RACE) 

adsl %>%
  count(SEX, TRT01P, RACE)
```

```{r, message = F, eval = T, echo = F}
#for multiple variables
table(adsl$SEX, adsl$TRT01P, adsl$RACE) 

adsl %>%
  count(SEX, TRT01P, RACE) %>%
  kbl() %>%
  kable_classic_2(full_width = F)
```

### KEEP/DROP $\rightarrow$ select()

```{sas, eval = F}
DATA adsl2;
  SET adsl;
  KEEP subjidn; *or drop subjidn;
RUN;
```

In R we use the `select()` function from the tidyverse package. 
```{r, eval = F}
# to keep (only) SUBJID
adsl %>%
  select(SUBJID) %>%
  head() # keep only first 6 rows

# to keep all variables except SUBJID
adsl %>%
  select(-SUBJID) %>%
  head()
```

```{r, eval = T, echo = F}
# to keep (only) SUBJID
adsl %>%
  select(SUBJID) %>%
  head() %>%
  kbl() %>%
  kable_classic_2(full_width = F)

# to keep all variables except SUBJID
adsl %>%
  select(-SUBJID) %>%
  head() %>%
  kbl() %>%
  kable_classic_2(full_width = F)
```

### Subsetting data (WHERE/IF) $\rightarrow$ filter()

```{sas, eval = F}
DATA female;
  SET adsl;
  WHERE SEX = 'F';
RUN;
```

```{r, eval = F}
adsl %>%
  filter(SEX == "F") %>%
  head() 

# multiple conditions:
adsl %>%
  filter(SEX == "F" & AGE > 70) %>%
  head() 
```

```{r, eval = T, echo = F}
adsl %>%
  filter(SEX == "F") %>%
  head() %>%
  kbl() %>%
   kable_classic_2(full_width = F)

# multiple conditions:
adsl %>%
  filter(SEX == "F" & AGE > 70) %>%
  head() %>%
  kbl() %>%
   kable_classic_2(full_width = F)
```

### Sorting data

```{sas, eval = F}
PROC SORT data=adsl out=adsl_sort_age;
  BY AGE; 
RUN;
```

In R we use the `arrange()` function from tidyverse: 

```{r, warning = F, message = F, eval = F}
# ascending
adsl %>%
  arrange(AGE)%>%
  head() 

# descending
adsl %>%
  arrange(-AGE)%>%
  head() 
```

```{r, warning = F, message = F, echo = F, eval = T}
# ascending
adsl %>%
  arrange(AGE)%>%
  head() %>%
  kbl() %>%
  kable_classic_2(full_width = F)

# descending
adsl %>%
  arrange(-AGE)%>%
  head() %>%
  kbl() %>%
  kable_classic_2(full_width = F)
  
```

### Creating new variables 
```{sas, eval = F}
DATA adsl;
  SET adsl;
  length AGEGR1 $20.; * length function in R means something different
  IF age > 50 then AGEGR1 = ‘> 50 years old’;
  ELSE if age <= 50 then AGEGR1 = ‘<= 50 years old’;
run;
```

In R we use the `mutate()` function:
```{r, eval = F}
adsl %>%
  mutate(AGEGR1 = case_when(
    AGE > 50 ~"> 50 years old",
    AGE <= 50 ~"<= 50 years old",
  )) %>%
  head()
```

```{r, eval = T, echo = F}
adsl %>%
  mutate(AGEGR1 = case_when(
    AGE > 50 ~"> 50 years old",
    AGE <= 50 ~"<= 50 years old",
  )) %>%
  head() %>%
  kbl() %>%
  kable_classic_2(full_width = F)
```

### Handling of missing values

Missing values in SAS and R:

* SAS 
  + Missing value is a blank/a single decimal point for character/numeric variables, see details in [Missing Values in SAS](https://documentation.sas.com/doc/en/pgmsascdc/9.4_3.5/lrcon/p175x77t7k6kggn1io94yedqagl3.htm);
  
* R
  + Missing data in R appears as `NA`. `NA` is not a string nor a numeric value, but an indicator of missingness, see detailed examples in [Advanced R #3.2.3 Missing values](https://adv-r.hadley.nz/vectors-chap.html#missing-values);
  + `NA` and `""` are different in R: `""` is a blank string, while `NA` is missing;
  + "Missing values (`NA`) and `NaN` values are regarded as non-comparable even to themselves, so comparisons involving them will always result in `NA`" 
  + `admiral::convert_blanks_to_na()` can turn SAS blank strings into proper R `NA` (from [admiral documentation](https://pharmaverse.github.io/admiral/reference/convert_blanks_to_na.html)).  

```{r, eval = T}
# Unlike SAS, space(s) != blank string in R, 
" " == ""
# is.na to check if it's missing
c(" ", "Y", NA_character_) %>% is.na()
# NA is non-comparable, result in NA
c(" ", "Y", NA_character_) != "Y"
```



Unexpected results might occur with "SAS habits", below are three examples to illustrate the different handling of missing values in R. 

#### Subsetting data

`!=` is not the same in R when it involves `NA`
```{r, eval = F}
# add one more row to the data with missing AGE and DISCONFL, select the first 3 rows
adsl_na_example <- adsl %>% 
  add_row(SUBJID = "1", DISCONFL = NA_character_, .before = T) %>% 
  slice(1:3) %>% 
  select(SUBJID, DISCONFL)
# demo data
adsl_na_example
```

```{r, eval = T, echo = F}
# add one more row to the data with missing AGE and DISCONFL, select the first 3 rows
adsl_na_example <- adsl %>% 
  add_row(SUBJID = "1", DISCONFL = NA_character_, .before = T) %>% 
  slice(1:3) %>% 
  select(SUBJID, DISCONFL)
# demo data
adsl_na_example %>%
  kbl() %>%
  kable_classic_2(full_width = F)
```

Subset to `DISCONFL != "Y"`, only `DISCONFL = " "` is selected, not `NA` 
```{r, eval = F}
adsl_na_example %>% filter(DISCONFL != "Y")
```

```{r, eval = T, echo = F}
adsl_na_example %>% filter(DISCONFL != "Y") %>%
  kbl() %>%
  kable_classic_2(full_width = F)
```

To display `NA`, add `is.na(DISCONFL)` to the filter
```{r, eval = F}
# include is.na(DISCONFL) in the filter
adsl_na_example %>% filter(DISCONFL != "Y" | is.na(DISCONFL))
```

```{r, eval = T, echo = F}
adsl_na_example %>% filter(DISCONFL != "Y" | is.na(DISCONFL)) %>%
  kbl() %>%
  kable_classic_2(full_width = F)
```





#### Sorting data

In SAS, missing is treated as the smallest value in `proc sort`; while in R, `NA` is always at the bottom after ascending or descending sorting.

Create demo data
```{r, eval = F}
# add one more row to the data with missing AGE and DISCONFL 
# select the first 3 rows
adsl_na_sort <- adsl %>% 
  add_row(SUBJID = "1", AGE = NA_integer_, .before = T) %>% 
  slice(1:5) %>% 
  select(SUBJID, AGE)
```

```{r, eval = T, echo = F}
# add one more row to the data with missing AGE and DISCONFL 
# select the first 3 rows
adsl_na_sort <- adsl %>% 
  add_row(SUBJID = "1", AGE = NA_integer_, .before = T) %>% 
  slice(1:5) %>% 
  select(SUBJID, AGE) 
adsl_na_sort %>%
  kbl() %>%
  kable_classic_2(full_width = F)
```

Sort by ascending order of AGE
```{r, eval = F}
# ascending
adsl_na_sort %>%
  arrange(AGE)
```

```{r, eval = T, echo = F}
# ascending
adsl_na_sort %>%
  arrange(AGE) %>%
  kbl() %>%
  kable_classic_2(full_width = F)
```

Sort by descending order of AGE
```{r, eval = F}
# descending
adsl_na_sort %>%
  arrange(-AGE)
```

```{r, eval = T, echo = F}
# descending
adsl_na_sort %>%
  arrange(-AGE) %>%
  kbl() %>%
  kable_classic_2(full_width = F)
```

`NA` can also be placed at the top with below workaround
```{r, eval = F}
# to be consistent with SAS - NA at the top when ascending
adsl_na_sort %>%
  arrange(!is.na(AGE), AGE)
```

```{r, eval = T, echo = F}
# to be consistent with SAS - NA at the top when ascending
adsl_na_sort %>%
  arrange(!is.na(AGE), AGE) %>%
  kbl() %>%
  kable_classic_2(full_width = F)
```


#### Creating new variables

When AGE is missing, AGEGR1 is set to '<= 50 years old' in SAS; while in R it is set to `NA_character_`.
```{r, eval = F}
adsl_na_sort %>%
  mutate(AGEGR1 = case_when(
    AGE > 50 ~"> 50 years old",
    AGE <= 50 ~"<= 50 years old",
  )) 
```

```{r, eval = T, echo = F}
adsl_na_sort %>%
  mutate(AGEGR1 = case_when(
    AGE > 50 ~"> 50 years old",
    AGE <= 50 ~"<= 50 years old",
  )) %>%
  kbl() %>%
  kable_classic_2(full_width = F) 
```

### Merging Data

We create another data set with weight information for some of the subjects: 
```{r}
SUBJID <- sample(adsl$SUBJID, 100, replace = FALSE)
WEIGHT <- sample(50:100, 100, replace = TRUE)

adsl_weight <- data.frame(SUBJID, WEIGHT)
```

Now, we would like to join the weight column to our adsl data set. 

In SAS, we use the merge function: 
```{sas, eval = F}
* inner join;
data adsl_merge;
merge adsl(in = inadsl) adsl_weight(in = inweight);
by subjidn;
if inadsl and inweight; * inner join in SAS
run;

* outer join;
data adsl_merge_all;
merge adsl(in = inadsl) adsl_weight(in = inweight);
by subjidn;
if inadsl or inweight; * default in SAS
run;

* left join;
data adsl_merge_left;
merge adsl(in = inadsl) adsl_weight(in = inweight);
by subjidn;
if inadsl;
run;
```

In R, we use again the tidyverse package: 

```{r, eval = F}
# inner join
adsl%>%
  inner_join(adsl_weight, by = "SUBJID") 

# outer join
adsl %>%
  full_join(adsl_weight, by = "SUBJID")

# left join
adsl %>%
  left_join(adsl_weight, by = "SUBJID")
```


### Concatenating Data

To demonstrate stacking or concatenating data sets, we can first split adsl into two parts, each having 5 records with the same columns in common.

```{r, eval = F}
# first 5 records of adsl
adsl_1 <- adsl[c(1:5),]

# the next 5 records of adsl
adsl_2 <- adsl[c(6:10),]
```

In SAS, you could use the set command:
```{sas, eval  = F}
DATA adsl_stacked;
 SET adsl_1 adsl_2;
RUN;
```

In R, you may accomplish this like so:
```{r, eval = F}
# tidyverse method - bind_rows()
adsl_stacked <- adsl_1 %>%
  bind_rows(adsl_2) 
```