---
title: A reactive sliced 3D surface response
date : 2013-03-15
--- &lead

```{r setup0, echo=FALSE}
opts_chunk$set(fig.path="assets/fig/rsmShiny", tidy=FALSE)
```

In [<u>my previous article</u>](http://stla.github.io/stlapblog/posts/rgl_knitr.html) I showed an interactive 3D surface response fitted from a model with two continous predictors. But when there is more than two continuous predictors, since we can use only two predictors at time in the image, we can only show a surface plot depending on the fixed values of the other predictors. Here we will build a `Shiny` application allowing to visualize a surface plot *reactive* to the values of the other predictors.

For our illustration we use the (coded) dataset `heli` available in the `rsm` package. 
The variables are coded variables `x1`, `x2`, `x3`, `x4`, which respectively are: linear functions of wing area
`A`, wing length ratio `R`, body width `W`, and body length `L`. The experiment was run in two
blocks, and the response variable is `ave`, the average flight time in seconds.

```{r}
library(rsm)
print(heli)
print(heli, decode=FALSE)
```


A second-order response-surface model for these data is fitted using
```{r, message=FALSE}
heli.rsm <- rsm(ave ~ block + SO(x1,x2,x3,x4), data = heli) 
```

The fitted surface response is shown below as a function of `x3` and `x4` with 
the values of `x1` and `x2` both fixed at $0$:  

```{r}
persp(heli.rsm, ~x3+x4, at=list(x1=0,x2=0), col = rainbow(50), contours = "colors")
```

Now we build a `Shiny` application which does the same plot with *reactive* values of `x1` and `x2`. 

**ui.R file** defining the user interface:
```{r, eval=FALSE}
library(shiny)
library(rsm)

shinyUI(pageWithSidebar(
  
  # Application title
  headerPanel("Sliced 3D plot"),
  
  # Sidebar with two slider inputs
  sidebarPanel(
    sliderInput("Aslice", 
                "A:", 
                min = 11.8, 
                max = 13, 
                value = 12.4),
    sliderInput("Rslice", 
               "R:", 
                min = 2.26, 
                max = 2.78, 
                value = 2.52)
  ),
  
  # Show plot 
  mainPanel(
    plotOutput("surfPlot")
  )
))
```

**server.R file** drawing the surface:
```{r, eval=FALSE}
library(shiny)
library(rsm)

heli.rsm <- rsm(ave ~ block + SO(x1,x2,x3,x4), data = heli)

shinyServer(function(input, output){
  output$surfPlot <- renderPlot({
    x1 <- (input$Aslice - 12.4)/0.6
    x2 <- (input$Rslice - 2.52)/0.26
    persp(heli.rsm, ~x3+x4, at=list(x1=x1,x2=x2), col = rainbow(50), contours = "colors")
    })
})
```
This `Shiny` application you see below is actually hosted at my  <s>*Shiny beta hosting account*</s> *ShinyApps hosting account* and 
is accessible at [https://stla.shinyapps.io/3Dsliced](https://stla.shinyapps.io/3Dsliced). 
I have embedded it here by typing the following html code: 

```
<iframe src="https://stla.shinyapps.io/3Dsliced" style="border: none; width: 1000px; height: 500px"></iframe>
```

<iframe src="https://stla.shinyapps.io/3Dsliced" style="border: none; width: 1000px; height: 500px"></iframe>



An interactive and reactive 3D plot ? 
---------------------------------------------

Similarly to interactive plots, reactive 3D plots embedded into a html report are great. Now, how to embed an interactive *and* reactive 3D plot ? 
There is [an open discussion](https://groups.google.com/forum/#!msg/shiny-discuss/VV-vw3VFpj0/bszBuHkgtS8J) about this possibility in the Shiny Google group. It's just a matter of time. 
How could we be able to afford to press the R developers who freely provide these wonderful tools ?