1) I have done only a minimal amount of testing of this program on Windows machines. Two issues that I am aware of are that the labels are not rounding correctly and that it is difficult to move the regression points. The trick is to move the mouse a little bit to the right and below the point. I think this is because the windows/mac mouse coordinates might differ by a little bit.
2) The bar charts represent the Sum of Squares and Mean Squares. In conjunction with the ANOVA table, I think that this configuration will show the connection between regression and ANOVA very well. Also, it shows the importance of sample size. That is, it shows clearly that as sample size increases, the MSerror is going to be reduced considerably becuse it is the SSerror divided by DFerror.
3) I have left the interface a little sloppy as I wanted to get more input as to how we want it to look, what information we want to have on the applet etc.
4) At this point, we should also consider the format we want to use for the applet. All of the other modules have had the applet inline w/ the tutorial. I was thinking, that it might be nice to have the applet appear in a new window. We could use a left hand side frame that has buttons such as "show applet", "show data", etc. so that when it is appropriate the user can bring up the program. We should think in terms of the tutorial, what we want the student to get out of it, and how we want to use it.
5) The applet has been double-buffered for smooth graphic rendering. The downside of this is that it will take a little work to make it run as an application. Before I double-buffered the applet, it was configured to work both as an application and an applet. It was really cool. The applet was an .exe file that by double clicking would run just like any other application on your computer. It would be very nice to have the applet as applications which teachers, students, whoever, could download. That way, one would not need to run the applet in a web browser making it easier and quicker to use as a class room demonstration or whatever. As we talk about applications, it might be interesting to recompile all of the applets into application(s) that can be distributed.
6) I've added a data editor (rudimentary at this stage) to the applet which can be brought up by clicking on the Data Editor button. The reason I felt this was useful was that when I was working with a student in the lab, I brought up the applet but felt very constrained by not being able to actually say what the values of the data were, and I also did not have a quick way to configure the data except by dragging the data points around. The data editor is not fault proof, for example, it only holds 15 cases right now even though the applet will hold much more than that.
7) The applet is written in Java version 1.1.8, which means that it will NOT
run on Netscape communicator for the Mac! It will run on Windows communicator
and IE, as well as IE 5 for the mac. Netscape 6 is in preview release for the
mac right now, and although I haven't tested it on that broswer, I expect that
it should run fine. I choose to write the applet in a newer version of Java
because the old system had a very klunky way of designing a user interface.
In fact, I can't believe I wrote user interfaces with the old system because
it was so tedious. Basically, for the data editor what would have taken at a
minimum of 100 lines of code I accomplished in 4. The new system is also a lot
more error prone because so much less code is needed. Also, this applet was
written using the Abstract Windowing Toolkit. The new model of java supports
components called Swing which look even better than AWT compoents. I'll have
to look into it more (I'm taking a class on the subject), but it might be interesting
to make this a swing applet.8
8) choose 'view page source' to view the applets tag. This will show you the parameters that we can adjust to configure the applet for every web page that it loads on.
As always, I appreciate your feedback on the applet, Mike
1) The unwanted resizing from dragging data points was removed
2) I've added some rough draft features. First, two buttons which zoom the scaling factor of the data points have been added. This is posible feature, which would need more work to smooth out, but
put it here mostly for demonstration. A bar graph representing SSerr and SSpred has been added as an alternative to the pie chart. Neither chart effectively handles the situation of the user
dragging the regression line to a point where r^2 becomes > 1.
3) The labels on the graph are now working
4) There have been a number of parameter tags added to the webpage so that the applet can be configured by the page which calls it.
5) Lines which connect to the x and y axes have been added. Clicking on a data point shows the line.
the regression line will highlight it. The regression line can then
be dragged with the mouse down.
The applet reads N then X and Y.
If N is greater than 0, it skips X and Y and generates N random numbers.
If X and Y are set, it uses these values as its initial data points.
If neither params are set or read correctly, the applet defaults to @10
Healy, M. R., Berger, D. E., Aberson, C. E., & Romero, V. L. (2000). A Java Applet for Teaching Correlation and Regression.
Last Updated 3/1/2000