Differences

This shows you the differences between two versions of the page.

--- beta_coefficients [2020/12/09 09:34] – [e.g.] hkimscil
+++ beta_coefficients [2026/06/14 23:44] (current) – hkimscil
@@ Line 9: / Line 9: @@
 <code>
 # import test score data "tests_cor.csv"
-tests <- read.csv("http://commres.net/wiki/_media/r/tests_cor.csv")
+tests <- read.csv("http://commres.net/_media/r/tests_cor.csv")
 colnames(tests) <- c("ser", "sat", "clep", "gpa")
 tests <- subset(tests, select=c("sat", "clep", "gpa"))
@@ Line 88: / Line 88: @@
 <code>
 # get marketing data
-marketing <- read.csv("http://commres.net/wiki/_media/marketing_from_datarium.csv")
+marketing <- read.csv("http://commres.net/_media/marketing_from_datarium.csv")
 head(marketing)
 # note that I need - X to get rid of X column in the marketing data
@@ Line 96: / Line 96: @@
 <code>
-> marketing <- read.csv("http://commres.net/wiki/_media/marketing_from_datarium.csv")
+> marketing <- read.csv("http://commres.net/_media/marketing_from_datarium.csv")
 > head(marketing)
   X youtube facebook newspaper sales
@@ Line 149: / Line 149: @@
 >
 </code>
+These beta coefficients also can be got from the coefficents from standardized data.
+<code>
+mod.formula <- sales ~ youtube + facebook + newspaper
+all.vars(mod.formula)
+marketing.temp <- sapply(marketing[ , all.vars(mod.formula)], scale)
+head(marketing.temp)
+mod.scaled <- lm(sales ~ ., data=marketing.scaled)
+head(marketing.scaled)
+coefficients(mod.scaled)
+</code>
+<code>> mod.formula <- sales ~ youtube + facebook + newspaper
+> all.vars(mod.formula)
+[1] "sales"     "youtube"   "facebook"  "newspaper"
+> marketing.temp <- sapply(marketing[ , all.vars(mod.formula)], scale)
+> head(marketing.temp)
+          sales     youtube   facebook newspaper
+[1,]  1.5481681  0.96742460  0.9790656 1.7744925
+[2,] -0.6943038 -1.19437904  1.0800974 0.6679027
+[3,] -0.9051345 -1.51235985  1.5246374 1.7790842
+[4,]  0.8581768  0.05191939  1.2148065 1.2831850
+[5,] -0.2151431  0.39319551 -0.8395070 1.2785934
+[6,] -1.3076295 -1.61136487  1.7267010 2.0408088
+> mod.scaled <- lm(sales ~ ., data=marketing.scaled)
+> head(marketing.scaled)
+       sales     youtube   facebook newspaper
+  1.5481681  0.96742460  0.9790656 1.7744925
+-0.6943038 -1.19437904  1.0800974 0.6679027
+-0.9051345 -1.51235985  1.5246374 1.7790842
+  0.8581768  0.05191939  1.2148065 1.2831850
+-0.2151431  0.39319551 -0.8395070 1.2785934
+-1.3076295 -1.61136487  1.7267010 2.0408088
+> coefficients(mod.scaled)
+  (Intercept)       youtube      facebook     newspaper
+-5.034110e-16  7.530659e-01  5.364815e-01 -4.330686e-03
+>
+> </code>
+check out that
+''lm.beta(mod) == coefficients(mod.scaled)''
+and
+베타를 구하고 나면 서로의 계수값을 절대비교할 수 있다.
+''lm.beta(mod)''의 아웃풋을 보고
+''youtube'', ''facebook'', and ''newspaper'' 순으로 설명력을 갖는다고 말할 수 있다 (혹은 newspaper를 분석에서 제외하고 다시 분석하여 둘만의 설명력을 보는 것도 방법이다.