Email
Archive for Project SAIL (2005)
|
Jessica |
|
More options |
Mar 13 |
Hello Java people
I hope this e-mail gets to
you all I’ve never actually used this mailing list before.
Here are the websites I
found in regards to the ideas we came up with last Friday.
Manually finding a linear
regression
http://people.hofstra.edu/faculty/Stefan_Waner/RealWorld/newgraph/regressionframes.html
negative: this program
doesn’t recalculate the line and graph until you tell it to; also it is
equipped to do additional types of regressions that we aren’t interested in;
also the column ‘predicated value’ is vague and you don’t understand where the
number it gives you comes from
Good combination of an
interactive graph and a visual for SSres (boxes)
http://standards.nctm.org/document/eexamples/chap7/7.4/
negative: won’t provide you with equation
for line of best fit
I didn’t find any valuable
websites that took you step by step through the process of finding the linear
regression. Either you had to pay to subscribe to access the program or the
instructions were specific to running a LinReg on a calculator.
Additional ideas:
Combine
entering data points and outputting an actual graph that reflects them into one
program
Link to LinReg program on TI83 (This
step is equivalent to the ‘r’ value provided by your calculator)
For visual of SSres
automatically display area of each box inside the box as well as the sum
of all the areas as the line is moved
Be
concise with instructions. Many of the programs I found had several paragraphs
of instructions that could have been said more concisely.
Incorporate graphs of x and y mean (and their intersection at y hat) into our
program. Consider building the graph as we go. Aka,
plot x mean. Plot y mean. Y hat must go through their intersection at this
point. Plot a y hat that allows the student to play with the slope to see the
changes in SSres and r but remains fixed
at the point (x mean, y mean)
See
everyone at
~Jess Kuntz~
|
Ariela Steif |
|
More options |
Mar 13 |
Hey everyone,
I got Jess's email...and decided to send out my own.
http://www.shodor.org/UNChem/math/lls/leastsq.html
This one you can enter data points and it gives you
the graph, best
fit line and equation, and r. Unfortunately, it doesn't give the
squares and SSdev and SSres,
which are kind of important.
http://people.hofstra.edu/faculty/Stefan_Waner/RealWorld/calctopic1/regression.html
This last one doesn't have an interactive visuals, but
it has some
charts and graphs, and it explains lin reg pretty well.
I also found Jess's second one, and it was the best one I could find out there.
Must go back to APUSH research paper. See you all
tomorrow.
Ariela
|
Andrew Owens |
|
More options |
Mar 13 |
Hi,
I'll likely bring in a crude drawing of the demonstration I have in
mind for tomorrow (or something like that).
Other linear-regression demonstrations:
http://www.math.csusb.edu/faculty/stanton/probstat/regression.html
- The demo takes place in a Java applet window rather than
an applet
inline in the webpage. This is definitely preferable to the inline
applet in this specific example as it allows the user to (relatively)
easily adjust the window's dimensions to better convey its data.
- The tabs seem to be a decent way of separating the
information,
though if the demonstration had anything more to it, there would
probably need to be a better way to separate the data.
- There really isn't a way to examine any given point – it
would be
nice if there was a list of every point and the information associated
with it. As it is now, there's just a graphical model that shows
every residual; the user gets a rough idea of what's going on, but it
would be helpful if he or she could look at a given point and
understand its deviation, distance from the residual line, etc.
- It's nice to be able to plot the points on the graph and
not enter
each individual point
http://www.math.csusb.edu/faculty/stanton/m262/regress/regress.html
- It's difficult to get an understanding of residuals'
relationship to
data simply from this demonstration because the residuals are in a
separate pane
- Again, it would be effective for the user to view each
point
individually (possibly in a table)
http://www.stat.uiuc.edu/courses/stat100/java/guess/PPApplet.html
- It's noteworthy that this demonstration allows you to
work
"backwards" – you can specify a target correlation and then allow the
applet to plot points that move the r value towards your specified
goal. This feature helps demonstrate correlation's relation to data
more effectively.
- You can specify whether or not the regression line
should be drawn
by simply checking or clearing a checkbox.
- It seems odd that the Residuals button creates a new
window since
it's a hassle that adds little to the demonstration; this would be
better implemented as a checkbox that allows the user to dynamically
view residuals like the regression line.
Ideas:
- There should be a single pane that displays the graph
and points;
having more than one adds unnecessary confusion to the demonstration
- There should be a separate tab that contains a large
variety of
statistics. Each statistic will have a corresponding checkbox. If
the user checks a box, then the demonstration will "track" the
variable. Only "tracked" variables are displayed to the user.
One
teacher could hypothetically tell his or her students to track only
the things that are featured in Precalculus Chapter 8
(SSdev, SSres,
residuals, etc.) while another teacher could tell his or her students
to only track least squares and med-med information. This list would
provide a level of extendibility while successfully hiding data that's
irrelevant to a given lesson; a teacher could tailor the demonstration
to his or her given curriculum. The list, of course, would be limited
to linear-regression only and we'd have to be careful not to try to
implement too many things.
- The user should be able to choose at run-time what
features the
graph should display. This dynamic could be implemented,
possibly as
a checkbox that turns on or off residuals or the regression line
- There should be a pane adjacent to the graph that
displays data the
user has chosen to "track" and nothing else
- There would be, possibly, another tab or section (a tab
may leave it
too cluttered and unusable) that allows the user to view information
about individual coordinates. There would also likely be a table of
mutable coordinates so the teacher could give a list of coordinates to
use.
- There should be a way to reproduce graphs; possibly allow
users to
load a file describing the points (I'm not sure about how we'd
implement this – I'd consider an importable/exportable XML file that
describes the demonstration and possibly make the demonstration into a
stand-alone application though this is way out of the scope of this
task right now)
- It may be a good idea to have a feature that allows the
user to
"work backwords" by giving a correlation
and having the demonstration
plot random points that shift it towards this correlation.
Sorry if I was a little long-winded with that description...
Thanks,
Andrew Owens
|
Douglas Tremblay |
|
More options |
Mar 13 |
Hi Everyone, Sorry for the lateness
http://www.stat.sc.edu/~west/javahtml/Regression.html
- This applet allows for the user to add a single point to an already
existing scatter plot. It furthermore displays the line of best fit
and the correlation coefficient.
- The main educational benefit is found with the affects of outliers
to the LSR line and correlation coefficient.
- With limiting abilities due to the fact that you can only add one
point, this interactive program also only displays two statistics. It
would have much more value if the SSdev. SSres,
and other stats were
shown.
- Furthermore, this website is poorly constructed in my opinion. With
no visible boundaries and countless limitations, the program's design
prohibits the user gaining substantial educational value.
http://www.stat.uiuc.edu/courses/stat100/java/guess/PPApplet.html
- This interactive software allows the user to create a scatter plot
in the designated space and it calculates the r value and line of best
fit. Furthermore, this program creates a residual plot and a
histogram of residuals. This also carries the feature of creating
random points on the graph and having a target correlation
coefficient.
- There is a lot of educational value in this program. The residual
plot is very useful, as well as the ability to create random points
with a target correlation. It is also very user friendly easy to
understand.
- However, this program lacks key statistics and could have many more
features, such as the visual display of the deviations and residuals
on the scatter plot as well as information on individual points.
- A well constructed and user friendly design adds to the interactive
benefit on this applet.
http://statweb.calpoly.edu/chance/applets/LRApplet.html
- This applet creates a scatter plot with existing points and allows
the user to add points or change existing points to the plot and
reports the r value, line of best fit, the selected point's position
and residual, and the SSE.
- There are nice features such as an expanding plane and the analysis
of individual points. It is also very simple and easy to use.
- Some problems include its limitation on the individual data as well
as statistics about the whole plot. It is best used to show the
information of individual points.
Ideas
- An interactive graph like the examples above and the idea we
discussed in class, with the main goal of understanding the
relationship within the graph for SSdev and SSres. This means a
visual display of both and a new feature of a second and third plane
that shows all the squares visually, then showing how they are summed
up. Additionally, the ability to move the line of best fit will also
affect the squares and consequently affect the two other graphs. The
math behind correlation coefficient will also be shown to enforce the
user's understanding.
- Another idea is to create guide through an example where the user
creates the points and must answer questions about graphs. This will
allow for a greater understanding about the graphical aspects of
linear regression and the visual process of the correlation
coefficient and the line of best fit.
Sorry Again,
Doug Tremblay
|
Andrew Owens |
|
More options |
Mar 14 |
Hi,
If anyone (especially the people that didn't submit an e-mail last
night) has any additional ideas or clarifications, please send them to
me now so I can compile them into a single list by tomorrow.
Tomorrow I plan on sending an e-mail to each person individually with
specific questions about his or her ideas. It would be helpful if
everyone could check their inboxes tomorrow night. Tell me if you
have any other suggestions.
Thanks,
Andrew Owens
|
|
|
|
More options |
Mar 14 |
|
Ariela
Steif Hey Andrew,
The excel sheet looks great, but I had another idea that was briefly
tossed around a few days ago -- interactive quizes so
the user can
check their understanding.
Thanks,
Ariela
|
|
|
More options |
Mar 14 |
I'm really sorry this got out so late, I just noticed that all of the
incoming mail was going to my junk folder and I just realized this, I'm really
sorry. Hopefully it doesn't cause anyone problems.
http://www.math.csusb.edu/faculty/stanton/m262/regress/regress.html
This
one is good, although it doesn't give the actual points.
http://www.mste.uiuc.edu/activity/regression/
This one is good too, but it doesn't really help in
figuring out what's
going on.
http://statweb.calpoly.edu/chance/applets/LRApplet.html
I like this one more than the other two, as it gives the data points and
other necessary information, although it still doesn't help with the math.
http://www.duxbury.com/authors/mcclellandg/tiein/johnson/reg.htm
This one is neat because it shows all of the little
box things, but once
again no math and you can't select the points.
http://www.stat.sc.edu/~west/javahtml/Regression.html
All this one shows is the effect that one point can
have on the whole line.
http://www.stattucino.com/berrie/dsl/regression/regression.html
This one is like the other ones, just a different
presentation.
-None of the websites show any math.
-None of
the websites have any tables of values for x, y, or anything else.
|
Andrew Owens |
|
More options |
Mar 14 |
Hi,
I've reproduced Ariela's message below. In the
future, send all SAIL
discussions to ml-math-java-2004@kosbie.net
so that everyone can view
them.
>> The excel sheet looks great, but I had another idea that was briefly
>> tossed around a few days ago -- interactive quizes
so the user can
>> check their understanding.
I've added your suggestion to the list. Would it be possible for you
to elaborate on this idea? Would this quiz be integrated into an
exploration or would it simply be a substitute for a paper quiz but
with interactive elements? This sounds like a
very good idea and it
would be helpful if you developed it further.
Does anyone else have any suggestions or variations on this idea?
(There's no need to debate them or anything, we'll do that later)
Oh, and someone please correct me if I'm giving bad advice.
Thanks,
Andrew Owens
|
|
|
|
More options |
Mar 14 |
|
Sorry, I actually meant to send it out to the mailing list, but I must
have just hit reply and sent it to Andrew.
As far as interactive quizes, I think the most
effective way to use
them is for it to be a part of the exploration, but not while the user
can actually look at the graph (so it tests what they remember from
the graph). If we end up having multiple graphs that only show one or
two concepts at a time (like have one graph that does y hat and
displayes best fit line with the points and explains
what y hat and
line are, and then another graph that has SSres and
explains the
boxes. In addition to these we could have one main graph that can
always be accessed but doesn't explain the individual areas so it
doesn't get cluttered. Btw, Andrew, this might be another thing to add
to the list.) Sorry for the massive tangent. Anyways, after each graph
there can be a quiz that tests solely what was on that one graph, and
then at the end perhaps a "test".
Hope this clarifies things, please let me know if any of it was confusing.
Ariela
|
Andrew Owens |
|
More options |
|
Mar
14Hi,
I've attached a matrix containing everyone's ideas. I'll talk with
each person individually about his or her ideas tomorrow through
e-mail and then I'll refine this list. Tell me what you think. (the
formatting is rather bad, I think, sorry).
Thanks,
Andrew Owens
|
|
|
|
|
More options |
Mar 15 |
|
Hi Andrew (et al).
First: excellent work, everyone! Keep up the good work!
As for the spreadsheet: good start. I'd like it to be formatted as
follows:
* each ROW is a FEATURE (like you have now)
* each COLUMN is a WEBSITE
* the FIRST row is the URL of the website
* the FIRST column is the feature description (like you have now)
* each VALUE is a measure of that website for that feature
(0 = not present, 1 = present but bad, 2 = ok, 3 = great)
* the LAST row is the SUM of all the scores for the website
* then SORT the table based on the sums, best-to-worst, left-to-right
Thanks.
DK
|
Andrew Owens |
|
More options |
|
Hi,
I've attached our current matrix according to Mr. Kosbie's
specifications, but it's rather crude right now. I've listed each
person's ideas in the first column, and I also listed the URL of most
of the sites that people suggested in the top row. (I excluded
several redundant or uninteresting submissions, but I'll add other
sites if someone requests them). I then ranked each site in terms of
the feature list. If you have any other questions regarding the
spreadsheet, take a look at Mr. Kosbie's e-mail or
send an e-mail to
the mailing list.
Now that we have this matrix, we need to revise it. Take a look at
each site, examining each feature implemented. In an e-mail to the
mailing-list, note which features seemed most effective, what
characteristics negatively impacted the demonstrations, how the sites
successfully implemented features, and any new or innovative ideas
you'd like to share.
Ideally, I'd like this e-mail by tomorrow night, but given that the
bluebook exam is on Thursday, I understand this may not be possible
for everyone. If you're really unable to complete this assignment by
tomorrow night, then please have it submitted by Thursday afternoon at
the latest.
With everyone's opinions, I'll revise the matrix by Friday.
If there are any suggestions, corrections, or comments, please send an
e-mail to the mailing-list.
Thanks,
Andrew Owens
|
|
Jessica |
|
More options |
Mar 15 |
In looking over the spreadsheet you made Andrew, I was confused by
several things. First,
What are visual boundaries
for applet?
Also,
What is a third pane?
And,
What are tabs?
In addition to that,
program ‘E’ doesn’t work for me. The box where the interactive graph ought to
be stays gray. Anyone else had this problem?
I’ll e-mail you my review
and response by tomorrow night. Thanks for the help!
Jess~
|
Andrew Owens |
|
More options |
Mar 15 |
Hi,
Thanks for the response. Sorry, I was unclear on a few things.
>> What are visual boundaries for applet?
>> What is a third pane?
Sorry, I probably should have cleared this up. When I recorded
people's feature suggestions, I accidentally took away the context
they were presented in a few times. I believe Doug mentioned that
http://www.stat.sc.edu/~west/javahtml/Regression.html
lacked clear
borders around its graph. This lack of boundaries presented a few
usability issues (specifically, it's possible to plot a point in the
-x range or even on top of the text near the bottom, though this isn't
clear due to the lack of borders).
As for the "third pane," I was referring to a suggestion I made
earlier to display all the statistics in a window. This was to be the
third display in a series, but out of context I agree that it really
doesn't make much sense and should probably be clarified or removed.
>> In addition to that, program 'E' doesn't work for me. The box where
the interactive graph
>> ought to be stays gray. Anyone else had this problem?
Yeah, it doesn't seem to load correctly in Internet Explorer and I
have no idea why (or it may be some other problem I'm overlooking).
If you really want to see it, you could install an alternative browser
like Mozilla Firefox (http://mozilla.org) - it seems to
work in it for
some reason. But you can probably skip program E otherwise since the
fact that it makes a separate window, using tabs to separate settings,
is really the only thing noteworthy about it.
>> What are tabs?
Good question. The only program that uses tabs, coincidentally, is
program E. A tab is a rectangular box that is used to logically
separate different sections in a window or dialog box. You've
probably seen them at one point or another.
Here's an example of tab use (bad tab use, specifically)
http://digilander.libero.it/chiediloapippo/Engineering/iarchitect/mfame.htm
Thanks for the response,
Andrew Owens
|
Andrew Owens |
|
More options |
Mar 16 |
Hi,
We have a new member! Steve Bradley will be joining SAIL and starting
work today.
Mr. Kosbie: His e-mail address is rotwatsb@hotmail.com - You'll have
to add him to the mailing-list if you haven't already.
Thanks,
Andrew Owens
|
Andrew Owens |
|
More options |
|
Hi,
Now that we have a list of features and a crude ranking of each site,
we need to determine the importance of each feature.
For every feature in the matrix, rank its importance on a scale from 1
to 10 (1 being unimportant, 10 being an absolute necessity). I'll
compile everyone's responses and use the average of these rankings to
rate each site on the matrix again. Also share any particularly
useful insights that you have.
In addition to the features we already have, we need to create some
truly unique ones. So, I'm asking everyone to submit at least one
"killer" feature.
Both the feature suggestions and the rankings will be due by tomorrow
evening. Send them together in a single e-mail to the mailing-list.
It would also be helpful if you could alert me of any revisions I need
to make on the matrix I've provided. Please tell me, for example, if
you think I ranked a site incorrectly or unfairly.
So, due tonight (or possibly tomorrow afternoon if you are absolutely too
busy):
- The feature evaluations that we discussed earlier
Due tomorrow evening:
a.) A rank for each feature
b.) One "killer" idea
If you have any questions or would like me to clarify something,
please send a message to the mailing list.
Steve: Because you're just joining us, I'm not going to require you
to rank each feature. You should still come up with a unique idea by
tomorrow, though. Also, you should examine the attached matrix and
visit each site to understand what's currently going on.
Thanks,
Andrew Owens
|
|
Jessica |
|
More options |
|
Hello everyone,
I have some specifics
about the list, but my general response was this: Overall, I think we need to
condense this list. There’s a good deal of overlapping going on; for example:
‘table of points’ and ‘plot points on graph or use table’ are listed as separate
features. Another example is feature 29, “Visual for SSres”
and feature 33, “SSres with area of given box and
total sum”. So, let’s group together like features and eliminate redundancy.
As I was looking through
the websites feature by feature I made specific notes on each (feature that is
and the websites that used it). It’s lengthy (given the multitude of features
to evaluate) so I’ll attach it.
If this is missing
something that you want Andrew, let me know what. Otherwise, good luck to
everyone tomorrow on the blue books! Try to sleep. Some.
Jess
|
Andrew Owens |
|
More options |
|
Hi,
Excellent work on the evaluations! (It was much more comprehensive
than I was expecting) You seem to have comprehensive knowledge of
these sites so, using the attached spreadsheet, could you evaluate
each site based on each feature on a scale from 0 to 3 as I did? (0 -
non-existent , 1 - partially there or poorly implemented, 2 - good
use, 3 - excellent)
I'd like this spreadsheet by tomorrow night. If you're unable to do
this, please tell me and I'll assign it to someone else since it's
definitely a substantial amount of work.
I'm looking for at least one more person to do this as well - please
volunteer if you can.
Thanks,
Andrew
|
|
|
|
More options |
|
||
|
Ahoy SAIL - ers (more specifically
Andrew), Sorry about the bad joke
again, but I can't get over it. I didn't know if you
wanted it as an Excel file or as a list, so I did both of them. Heres my list of features from one to
ten. 10 - Table of points |
|
||||
|
7 - More
verbose with math (less magic) |
|
||||
|
8 - Can
toggle Ssdev and Ssres on
and off |
|
||||
|
8 - Must
allow multiple points |
|
||||
|
5 -
Visible boundaries for applet (clear borders around the exploration |
|
||||
|
8 - Visual display of deviations/regression | |||||