I have a dataframe like so
wood <- read_csv("/Users/name/Desktop/AR/Exercise data-20201109/woodstrength.csv")
I select x and y
x <- wood %>% select(Conc)
y <- wood %>% select(Strength)
The relationship can be modeled with a polynomial of degree 2 so I do
m <- lm(y ~ poly(x, 2, raw = TRUE))
which returns
non-numerical argument for binary operator
but x looks like this
> x
# A tibble: 19 x 1
Conc
<dbl>
1 1
2 1.5
3 2
4 3
5 4
6 4.5
7 5
8 5.5
9 6
10 6.5
11 7
12 8
13 9
14 10
15 11
16 12
17 13
18 14
19 15
What am I doing wrong?
If you look at the help page (?poly) for poly :
poly(x, ..., degree = 1, coefs = NULL, raw = FALSE, simple = FALSE)
[...]
x, newdata: a numeric vector at which to evaluate the polynomial. ‘x’
can also be a matrix. Missing values are not allowed in ‘x’.
Your dataset is tibble, when you do select it keeps the object as a tibble:
wood = tibble(`Conc` = rnorm(10),'Strength'=rnorm(10))
x <- wood %>% select(Conc)
class(x)
[1] "tbl_df" "tbl" "data.frame"
You get that error because underneath the function, it applies something that expects a matrix or vector, but sees a list or data.frame or in your case a tibble, hence the error. You can see why calling the column out works:
class(wood[["Conc"]])
[1] "numeric"
To get it to turn into a numeric or vector, you can do:
x <- wood %>% pull(Conc)
y <- wood %>% pull(Strength)
m <- lm(y ~ poly(x, 2, raw = TRUE))
Or:
m <- lm(Strength ~ poly(Conc, 2, raw = TRUE),data=wood)