r - How Can I manually obtain predict() values from coef/model.matrix returns on linear model -

i having lot of trouble figuring out how find predicted values coefficients of models , model matrix. i'm hoping can help.

i have linear model 2 independent variables setting up. e.g.

data <- data.frame(d1,d2,d3) lm.data <- lm(d1~d2*d3,data) 

i can coefficient vector

co.data <- coef(lm.data) 

i can model matrix

mm.data <- model.matrix(lm.data) 

this can lost!! trying teach myself how can match values can when use predict(lm.data) coefficients. in other words, know predicted values of model design matrix , coefficients can calculated, after past 48 hours of working on this, have no idea.

any amazing.

you need know how linear model works. if formula d1 ~ d2 * d3 , they're numeric, predict (intercept) + (d2 coefficient)*x_d2 + (d3 coefficient)*x_d3 + (d2:d3 coefficient)*x_d2*x_d3 , give predicted d1.

here's reproducible example:

data(iris) m <- lm(sepal.length ~ petal.length * sepal.width, iris) co.data <- coef(m)  # we'll predict sepal length these petal lengths , sepal widths: x.pl <- runif(5, min=1, max=2) x.sw <- runif(5, min=2, max=5) y.predicted <-  predict(m, data.frame(petal.length=x.pl, sepal.width=x.sw))  #        1        2        3        4        5  # 5.379006 5.495907 5.296913 4.382487 5.131850  

now manually, let's @ coefficients:

co.data # intercept)             petal.length              sepal.width petal.length:sepal.width  # 1.40438275               0.71845958               0.84995691              -0.07701327  

according formula above:

y <- co.data[1] + co.data[2]*x.pl + co.data[3] * x.sw + co.data[4]*x.pl*x.sw # [1] 5.379006 5.495907 5.296913 4.382487 5.131850 

rather writing out manually can like:

# x matrix columns 1, petal length, sepal width, pl*sw # (matches order of co.data) x <- cbind(1, matrix(c(x.pl, x.sw, x.pl*x.sw), ncol=3)) x %*% co.data #          [,1] # [1,] 5.379006 # [2,] 5.495907 # [3,] 5.296913 # [4,] 4.382487 # [5,] 5.131850