R Regression - Revision history

Rb: /* Accessing Regression Output */

2015-08-29T08:56:21Z

‎Accessing Regression Output

2015-08-29T08:55:19Z

‎Accessing Regression Output

2015-08-29T08:43:32Z

‎Accessing Regression Output

2015-08-29T08:42:51Z

‎Accessing Regression Output

2015-07-09T19:42:52Z

‎A first example

2015-07-07T07:51:37Z

2015-07-06T20:54:15Z

2015-07-06T20:23:33Z

2015-04-14T14:26:16Z

‎Accessing Regression Output

2015-04-14T14:25:53Z

‎Accessing Regression Output

@@ Line 119: / Line 119: @@
      reg_ex1_sm<source enclose=none>$</source>coefficients[2,"Std. Error"]
-Only two things to mention here. The coefficients variable in the summary output is really a matrix (here with 3 rows and 4 columns) and yo can access individual elements by indicating which  the respective <code>[row,col]</code>. But instead of using row and column numbers you can also use the row's and column's names.
+Only two things to mention here. The coefficients variable in the summary output is really a matrix (here with 3 rows and 4 columns) and yo can access individual elements by indicating the respective <code>[row,col]</code>. But instead of using row and column numbers you can also use the row's and column's names.
 As you would often want to use values from this matrix, the kind people who developped R also devised a slightly shorter way to do this, using the <code>coef</code> method. This results in the following alternative ways to access the standard error to the exper coefficient estimate.

@@ Line 111: / Line 111: @@
      reg_ex1_sm<source enclose=none>$</source>coefficients[3,4]
      [1] 0.0002320238
 == Extra parameters to the <source enclose=none>lm()</source> function ==

@@ Line 91: / Line 91: @@
       [13] "model"
-If you compare the lists you can see that there is some overlap between the information contained in the two lists. In fact I find that the only list member that you would usually want to access from <source enclose=none>reg_ex1</source> directly is the <source enclose=none>"fitted.values"</source>. Most other commonly used statistics are available from the summary list.
+If you compare the lists you can see that there is some overlap between the information contained in the two lists. In fact I find that the only list member that you would usually want to access from <source enclose=none>reg_ex1</source> directly is the <source enclose=none>"fitted.values"</source>. Most other commonly used statistics are available from the summary list. I therefore make it a habit to save the summary as above.
 The question now remains how you access particular members of these lists if you wanted to use them further down in your code. There are a number of ways of accessing members of a list. This isn't really the place to discuss these in detail. [http://www.r-tutor.com/r-introduction/list/named-list-members This] is a good place to go if you want to find out more. I shall just demonstrate one way to do this.

@@ Line 91: / Line 91: @@
       [13] "model"
-If you compare the lists you can see that there is some overlap between the information contained in the two lists. In fact I find that only list member that you would usually want to access from <source enclose=none>reg_ex1</source> directly is the <source enclose=none>"fitted.values"</source>. Most other commonly used statistics are available from the summary list.
+If you compare the lists you can see that there is some overlap between the information contained in the two lists. In fact I find that the only list member that you would usually want to access from <source enclose=none>reg_ex1</source> directly is the <source enclose=none>"fitted.values"</source>. Most other commonly used statistics are available from the summary list.
 The question now remains how you access particular members of these lists if you wanted to use them further down in your code. There are a number of ways of accessing members of a list. This isn't really the place to discuss these in detail. [http://www.r-tutor.com/r-introduction/list/named-list-members This] is a good place to go if you want to find out more. I shall just demonstrate one way to do this.

@@ Line 13: / Line 13: @@
       # This is my first R regression!
       setwd("T:/ECLR/R/FirstSteps")              # This sets the working directory
-      mydata <- read.csv("mroz.csv")  # Opens mroz.csv from working directory
+      mydata <- read.csv("mroz.csv",na.strings = ".")  # Opens mroz.csv from working directory
-−
       # Run a regression
       <span style="color:#0000ff">reg_ex1 <- lm(lwage~exper+log(huswage),data=mydata)</span>

@@ Line 136: / Line 136: @@
 The term that follows <source enclose=none>subset =</source> is a vector of boolean variables (FALSE/TRUE variables). Here we are asking the question (of all our available observations) whether the value of the <source enclose=none>exper</source> is larger than 9. If it is the answer is TRUE and that observation is included in the subset, if it is not the answer is FALSE and that observation will be dropped from the regression.
 == Footnotes ==
   <references />

@@ Line 130: / Line 130: @@
 === Regression on subsamples ===
-Being able to easily repeat a regression on just a sub-sample of data is an important tool in the econometrician's toolbox. Fortunately this is done very easily in R. We use the additional parameter into the <source enclose=none>lm()</source> function, <source enclose=none>subset</source>. Let's see an example and then dicuss how it works:
+Being able to easily repeat a regression on just a sub-sample of data is an important tool in the econometrician's toolbox. Fortunately this is done very easily in R. We use the additional parameter into the <source enclose=none>lm()</source> function, <source enclose=none>subset</source>. Let's see an example and then discuss how it works:
       reg_ex1 <- lm(lwage~exper+log(huswage),data=mydata, subset = (exper > 9))
@@ Line 138: / Line 138: @@
 == Using Categorical/Factor variables in regressions ==
-as we discussed in the [[ R_Data|Data Section]] categorical variables are label factor variables in R.
+as we discussed in the [[ R_Data|Data Section]] categorical variables are labelled as factor variables in R. When using such categorical variables in regressions as explanatory variables we will use them in the form of dummy variables. Our practice dataset doesn't have any real factor variables (remember, wages and log(wages) were initially treated as factor variables due to the missing values.
 == Footnotes ==
   <references />

@@ Line 105: / Line 105: @@
 The version from the regression output reports the coefficients only, while the summary version also contains standard errors, t-statistics and p-values:
-     reg_ex1_sm$coefficients
+     reg_ex1_sm<source enclose=none>$</source>coefficients
                     Estimate  Std. Error  t value     Pr(>|t|)
      (Intercept)  0.53486640 0.139081935 3.845693 1.386203e-04
@@ Line 111: / Line 111: @@
      log(huswage) 0.23646570 0.063684482 3.713082 2.320238e-04
-     reg_ex1_sm$coefficients[3,4]
+     reg_ex1_sm<source enclose=none>$</source>coefficients[3,4]
      [1] 0.0002320238