Example of using the “Single-Sample Variance Test…” menu

First we will load the randtest package from which we will load the data. For it:

1. From the Rcmdr menu, choose the option “Tools” -> “Load package(s) …”
2. Search in the randtest dialog box and select it
3. Press accept

In the message window should appear [x] NOTE: Packages loaded: randtest. If randtest does not appear in the list, you must first install the package.

Next we will load a set of data, for example, “sweetpotato” through the following steps:

1. Choose from the Rcmdr menu: “Data” -> “Package data sets” -> “Read data set from attached package…”
2. Press twice on “randtests”
3. Click on “sweetpotato” and on “Accept”.

Rcmdr reply with the following text in the instruction tab (R Script)

data(sweetpotato, package = "randtests")
sweetpotato <- as.data.frame (sweetpotato)

In the message window it will appear [x] NOTE: The data set sweetpotato has 70 rows and 4 columns. and in the dialog box attached to the label “Data set:” “sweetpotato” will appear.

To build the confidence interval for sigma for the variable “yield”, select from the Rcmdr menu: “Statistics” -> “Variances” -> “Single-Sample Variance Test…” select “yield” and “OK”. Rcmdr responds with the following instruction in the instruction tab (R Script)

with(sweetpotato, sigma.test(yield[!is.na(yield)], alternative='two.sided', sigmasq=1.0, conf.level=0.95))

and in the output tab

with(sweetpotato, sigma.test(yield[!is.na(yield)], alternative='two.sided', sigmasq=1.0, conf.level=0.95))
#> 
#>  One sample Chi-squared test for variance
#> 
#> data:  yield[!is.na(yield)]
#> X-squared = 6514.8, df = 69, p-value < 2.2e-16
#> alternative hypothesis: true variance is not equal to 1
#> 95 percent confidence interval:
#>   69.41231 135.93966
#> sample estimates:
#> var of yield[!is.na(yield)] 
#>                    94.41731

The null hypothesis \(\sigma = 1\) has been rejected at a confidence level of 95%, obtaining a confidence interval at that level for \(\sigma\) of \((69.41231, 135.93966)\).

If desired, in the previous dialog you can specify:

1. the alternative hypothesis
2. the value of sigma under the null hypothesis
3. the level of trust.

Randomness test for a dichotomous variable

Within the menu “Statistics” -> “Non-parametric tests” -> “Randomness test for a factor of two levels…”, a new entry is provided to test the randomness of a factor-type variable with two levels. This option uses the runs.test function of the tseries package, although to avoid conflicts it has been renamed as twolevelfactor.runs.test.

data(AMSsurvey, package = "carData")

with (AMSsurvey, twolevelfactor.runs.test (sex))

with (AMSsurvey, twolevelfactor.runs.test (sex))
#> 
#>  Runs Test
#> 
#> data:  sex
#> Standard Normal = 4.5917, p-value = 4.397e-06
#> alternative hypothesis: two.sided

Randomness test for a numerical variable

Within the “Statistics” menu -> “Non-parametric tests” -> “Randomness test for a numeric variable …”, a new entry is provided to test the randomness of a numeric variable. This option uses the runs.test function of the randtest package, although to avoid conflicts it has been renamed as numeric.runs.test.

with (sweetpotato, numeric.runs.test (yield))

with (sweetpotato, numeric.runs.test (yield))
#> 
#>  Runs Test
#> 
#> data:  yield
#> statistic = -4.5751, runs = 17, n1 = 35, n2 = 35, n = 70, p-value =
#> 4.759e-06
#> alternative hypothesis: nonrandomness

Input data and predict

If you select “Input data and predict”, a new data set, as a data.frame, will be created and the editor will be invoked. Then you can entry the values of the predictor variable that you want to use for prediction, the values for non predictor variables are not required. When you close the data editor the predicted values for predicted variable are shown.

data(Chile, package="carData")

RegModel.1 <- lm(income~age, data=Chile)
summary(RegModel.1)

RegModel.1 <- lm(income~age, data=Chile)
summary(RegModel.1)
#> 
#> Call:
#> lm(formula = income ~ age, data = Chile)
#> 
#> Residuals:
#>    Min     1Q Median     3Q    Max 
#> -33175 -23653 -17457   1673 168847 
#> 
#> Coefficients:
#>             Estimate Std. Error t value Pr(>|t|)    
#> (Intercept) 37240.54    2176.34  17.112   <2e-16 ***
#> age           -86.96      52.68  -1.651   0.0989 .  
#> ---
#> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> 
#> Residual standard error: 39500 on 2599 degrees of freedom
#>   (99 observations deleted due to missingness)
#> Multiple R-squared:  0.001047,   Adjusted R-squared:  0.000663 
#> F-statistic: 2.725 on 1 and 2599 DF,  p-value: 0.09891

.data <- edit(Chile[0,])
.data
predict(RegModel.1, .data)
remove(.data)

.data <- edit(Chile[0,])
.data
#>   region population  sex age education income statusquo vote
#> 1   <NA>         NA <NA>  34      <NA>     NA        NA <NA>
#> 2   <NA>         NA <NA>  45      <NA>     NA        NA <NA>
predict(RegModel.1, .data)
#>        1        2 
#> 34283.74 33327.13
remove(.data)

Add predictions to existing dataset

If you select “Models” -> “Predict using active model” -> “Add predictions to existing dataset…” the predictions are added to the selected data set using the active model and the selected data set for the values of the explanatory variables. After selecting this option, the user can select an existing data set using dialog box.

If the data set does not provides the values for all predicting variables an error will occur and no predicted values will be provided.

Unlike the menu option “Add observation statistics to data…”, this option can be used with a different data set than the one used to construct the model, if that dataset provides the values for all the predictor variables.

Example of use of “Add predictions to existing dataset…” menu entry

Load data “Chile” as in the previous example.

To build a model select from Rcmdr menu: “Statistics” -> “Model fit” -> “Linear Regresion…” as “Response variable” select income and age as “Explanatory variables” and click on “OK”. Rcmdr reply with the following command in source pane (R Script)

RegModel.1 <- lm(income~age, data=Chile)
summary(RegModel.1)

and in the output box

RegModel.1 <- lm(income~age, data=Chile)
summary(RegModel.1)
#> 
#> Call:
#> lm(formula = income ~ age, data = Chile)
#> 
#> Residuals:
#>    Min     1Q Median     3Q    Max 
#> -33175 -23653 -17457   1673 168847 
#> 
#> Coefficients:
#>             Estimate Std. Error t value Pr(>|t|)    
#> (Intercept) 37240.54    2176.34  17.112   <2e-16 ***
#> age           -86.96      52.68  -1.651   0.0989 .  
#> ---
#> Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
#> 
#> Residual standard error: 39500 on 2599 degrees of freedom
#>   (99 observations deleted due to missingness)
#> Multiple R-squared:  0.001047,   Adjusted R-squared:  0.000663 
#> F-statistic: 2.725 on 1 and 2599 DF,  p-value: 0.09891

Note that the active model is set to RegModel.1 So if you want to predict the values for income for age data in Chile dataset. Select from Rcmdr menu: “Models” -> “Predict using active model” -> “Add predictions to existing dataset…”. In the dialog select a compatible dataset with the model. In this case select Chile. Rcmdr reply with the following command in source pane (R Script)

Chile$fitted.RegModel.1 <- predict(RegModel.1, Chile)

and in the output box

Chile$fitted.RegModel.1 <- predict(RegModel.1, Chile)

The predicted value of income has been saved as fitted.RegModel.1 in the selected dataset (Chile). You can see the added values using the button for visualizing the data set.