Recent Changes -

ComparisonWithThePackageRcmdr {Main}

I (Ian Fellows) was recently asked to compare Deducer with R Commander. This is my reply:

Let me start off by saying that I really like what Dr. Fox has done. He has created a great resource for beginners to get familiar with R. Rcmdr is designed as an interface for beginners, and it is Dr. Fox's hope that they will be "weened off of the dialogs" and eventually just use the command line. I have a bit of a different goal with Deducer. I want Deducer to be user friendly enough to be used by someone just starting out, yet flexible and powerful enough to increase the efficiency of expert users. It would be unfair to criticize Rcmdr for not serving the needs of advanced users, because it was never designed for this. That said, I think Deducer has a lot of advantages, both in ease of use, and increased efficiency.

Ease of use:

  1. Console integration: It is a bit awkward to have a separate console for the menus as Rcmdr has. Deducer is fully integrated into the usual windows console, though it is best used with the user friendly cross-platform JGR console.
  2. Data viewer: Beginners really like to be able to see their data. The data viewer goes a long way in helping them see what they are doing.
  3. Task based dialogs: Deducer has task based dialogs instead of function based ones. Rcmdr has separate dialogs for each R function it uses. For example, there are 6 menu items for loading data from a file (one for each loading function), whereas Deducer has one which picks the right function based on extension and user selection. As another example, two-sample tests in Deducer are all grouped into one dialog because they are all different ways of answering the same question.
  4. Help system: Deducer's online help system walks the user through the dialogs with screen shots and practical examples.
  5. Prettier interface: The toolkit that is used to make Rcmdr is somewhat limiting, and (IMHO) doesn't have the nicest look and feel.
  6. Drag and drop: Deducer implements drag and drop to and from the dialog lists.

Efficiency:

  1. Dialogs have memory: One of the constantly vexing things about Rcmdr is that the dialogs do not remember what options were selected the last time they were used. Data analysis is an iterative process, it is very rare that I specify exactly the right set of options the first time.
  2. Multiple datasets: Deducer can work with multiple data frames at the same time. No need to specify an "active set." Deducer automatically defaults the variable selector to the most recently used dataset, but this can be changed easily for each analysis.
  3. Multiple analyses: sometimes you'll want to do a t-test on 5 variables. In Rcmdr, you have to go through the dialog 5 times.

Both:

  1. Result tables: tests and analyses are formatted into easy to read tables. This is particularly apparent when comparing the two-programs' implementations of contingency tables.
  2. Analysis plots: Each analysis has at least one descriptive plot which helps users visualize their results, and identify potential problems such as outliers/non-normality.

Areas where Rcmdr is superior:

  1. Excel integration: Rcmdr can be integrated into Excel.
  2. Richer set of analyses: Rcmdr is more mature, and thus has a more rich set of analyses that have been programmed into it, including many excellent plug-ins.