ER and EED, Questions and Answers
If you have questions or comments, you can send us e-mail
ER Download page, J. Werner's Page (English), J. Werner's Page (Deutsch) , Dave Steppan's Homepage
Good News for the users of the French Language versions of Excel! See here for a solution of the encryption problem! Thanks to Bertrand for finding this!
Release 2.219 of Essential Regression (file date 04/01/01) with a bug fix in the code that runs when the user presses the Graph button directly on the Regression dialog. As one of our faithful and alert users (Thanks, Chris!) told us, the X and Y axis were reversed the first time the Graph button was used.
Not really news, but two remarks, based on e-mails we have been receiving:
-ER and EED should run in Excel 2000 (see Q&A #8, below)
-There seem to have been problems with the file download. We are considering creating a mirror download site so that interested users can be re-routed if needed. We will announce this as soon as the mirror site is established.
Ever since Win98 came out, we have been receiving complaints by Win98 users that the setup is not working properly. How to get around this problem is described below under Questions and Answers, #7.
I was told by several visitors of our web site that the feedback form was not working. Well, folks, what happened was that, after Yahoo bought Geocities, I had to change my member name. The feedback form still had my old name in it, that 's why it didn't work. It should work again now.
New Update, ER Version 2.218 (3/13/99), contains 'Near-Neighbor Lack of Fit'-Analysis. This allows for the estimation of pure error even without replicates in the data set.
We explain this type of analysis and also other changes made to ER since we wrote our Electronic Book here. For the future, we plan on publishing this 'addendum' as a Adobe Acrobat file similar to the Electronic Book.
We were recently made aware of a problem in the Central Composite Design module of EED for XL97. Basically, EED would create the same design irrespective of the type of CCD chosen. We fixed this problem that was caused by an oversight on our side when translating the XL95 code of EED into XL97 VBA.
The new version of EED is called 2.206 (dated 1/26/99) and is part of the updated ER/EED package (erzip97.zip) that can be downloaded from our web page ER Download page.
Even though we discussed it in our problem section below, users keep asking us about the error message:
As we discuss further below, this message occurs when the Solver add-in cannot be accessed by our software, either because it is not installed or because the path is not correct. Please refer to the problem section concerning the Solver below.
As some user might have noticed, ER issues an error message if you try to perform a regression with more variables than total degrees of freedom. If you received messages such as:
Main regression done, In residual calculations
Other residual calculations crashed
Subscript out of range
See comments below under 6.
This is, first of all, a thank you to all the people who have been trying ER and EED since we published the addins and the electronic book on the web. The response was actually much better than we initially expected, not having a lot of experience with publishing on the web. A special thanks to the users who gave us their feedback on technical issues or just their opinions. We keep trying to respond to your e-mail promptly. If somebody did not receive a response in the last two or three weeks, the reason might be that everybody has to take a vacation sometimes, and then things can get mixed up. If your question is still unanswered please try again.
A technical issue: we decided to stop supporting Excel 5.0. This version is quite old, and we cannot guarantee full compatibility anymore. We recommend using Excel for Win95 (=Excel 7.0) or higher (Excel 97 = Excel 8.0 and higher).
Some people have asked if they could use ER for non-linear modeling (logistic regression etc.). See comments below.
New Versions of ER and EED for Excel97 are now posted on our download page (Note: the book and tutorial do not include these new features yet, but we're working on it!)
ER, Version 2.217 (9/18/98) for Excel97:
-Added multiple prediction table in addition to predicting new values data point by data point. From 5 to 50 new data points can be defined and, based on the regression model, the new responses (including confidence limits) are calculated. In addition, the diagonal elements of the hat matrix are calculated and the user can directly see from the color of the cell if the new data point extrapolates the existing model.
-New buttons to add and remove all terms in a model in one step.
-Fixed problems with optimization of polynomial regression models. Fixed problems with intercept free regression.
EED, Version 2.205 (9/18/98) for Excel 97:
-Addressed problem with multiple response variables. EED can now accept more responses.
Questions and Answers
back to the News
1. Download Problems
We have received quite a number of complaints about the fact that downloading our software from the Geocities server can be a problem. Unfortunately, this is not a commercial venture and we set up on Geocities because at least it is free. Geocities can be difficult to get anything off of during the business hours in the USA. It seems to operate fine from 10PM to 9AM Eastern Standard Time (USA). We can only suggest that you download the software during those hours. Wehave tried it ourselves repeatedly between 8AM and 9AM EST and it worked fine. Our e-mail provider does not let us attach such large files. Please understand that we cannot send ER or EED as e-mail attachments.
The latest versions of Internet Explorer (not tested on Netscape) seem to resume from where a download was dropped. Sometimes restarting the download is all it takes to resume where it was cut off.
User 1 from the Netherlands:
..Even whith the 'solver.xla' installed the
error message you mention on your homepage keeps appearing...
..The dutch version of Excel doesn't have a 'solver.xla'. It is named after the dutch word for 'solver', being 'oplosser.xla', which may cause your program to be unable to find it...
My solution: I've copied 'oplosser.xla' and renamed it 'solver.xla'. Changing this in the menu however doesn't seem to be remembered by Excel. So I added solver.xla to the ER program group and start it manualy every time I want to use ER. It seems to work, ER then starts as it should.
User 2 from Italy:
... when I load ER22.xla ..., Excel prompts me "Path not found" (this is a translation from Italian)...
User 3 from the U.S.A.:
I tried to to download your Essential Regression Program. I am able to start the experimental design section properly, however, the Essential Regression Program gives me an error that says "Path not found"...
User 4 from the U.K.:
..I downloaded and installed Essential Regression. when I want to use ER22.xla I got the message "Path not found". .
User 5 from the U.S.A:
....I get the error message "compile error in hidden module: ModStartup"...
User 6 from Denmark:
...As I am working in Danish Excel, I also experienced initial difficulties in loading, since the "Library" folder is called "Bibltek" in Danish. However, I just created new folders "Library" and "Solver", corresponding to the path you specified, and copied solver.xla into "Solver". Works fine, so far ;).
Our Answer 1:
There is a known issue concerning the Solver-AddIn in Excel. In situations similar to the ones described above, start the Solver AddIn before starting ER (check the Solver AddIn in the list of AddIns under "Tools, AddIns" (English Version).
Make sure that the file "solver.xla" is in either of the following directories on your computer:
for the Excel 95 version: C:\MSOffice\Excel\library\solver\solver.xla
for the Excel 97 version: C:\Program Files\Microsoft Office\Office\library\solver\solver.xla
this file must be copied and renamed and loaded every time if you have a non-English version of Excel where the file name is different from "solver.xla".
Our Answer 2 (if you tried to start ER using one of the Desktop Icons):
Try starting ER from within Excel (File, Open..., "Er22.xla")
3. Problems related to Encryption and Passwords
There appears to be a problem in the French version of Excel. We received several e-mails from France such as:
...I tried to open the version downloaded (both the .xla file and the .xls example file) and I got a message from excel that there is a Password encrytion method in the file that is not supported in the French version...
Betrand showed us a way to circumvent this problem:
There is a way for french users to bypass this limitation in Excel 97 :
Go to (Win98) : "Demarrer" -> "Parametres" -> "Panneau de configuration" -> "Parametres regionnaux". TThen choose anything else than "Francais", for example "Francais (Belgique)". Go to "Symbole monetaire" and change "FB" to "F".
Now ER/EED should run
...I also found some differences between an SPSS output and an ER output in the correlation matrix...
This is most likely due to the use of (or the lack of) standardization of the variables, especially in models with quadratic, interaction, or higher order terms. Most "big" software packages standardize the linear variables by default. In ER you have to select this option.
5. ER and non-linear models (non-linear regression, logistic regression, etc.)
Non-liear regression (logistic etc.) is possible if you can linearize the model equation. If there is an exponential expression, you can form the logarithm etc. We have written something about this topic in our electronic book. Any model equation that can be transformed in a sum of expressions where the variables are separated (i.e., the partial derivative of one variable is independent of the others), can be used to perform a (multiple) linear regression. Problems can occur, however, with the error structure. For example if the original equation is
y = AexpB(x)+Error
then in the linearized equation
the error structure is different from the one in the original equation.
6. Regression Models with too many variables
If you try to do this, you will receive the following error message when running the regression:
Main regression done, In residual calculations
Other residual calculations crashed
Subscript out of range
Try reducing the regression model size by using less input variables or a simpler model, i.e., a linear model or higher order model with less cross-terms ("second order, no interactions" etc.). Also, you can try and start an "auto regression" and see what kind of model the software comes up with. Alternatively, you can perform a forward regression starting with no input variable at all.
7. Setup does not work (reported by Win98 users)
There seems to be a problem with our setup in WIn98, probably due to changes in the VB engine. However, for Win98 users, there is a way to get around setup and get started with ER and EED anyway:
After the first unzip of the downloaded files using WInzip, you'll get a bunch of files with an underscore "_" at the end.
Using Winzip, you can extract those files again into a single new directory named c:\eregress. You"ll have to rename them while extracting them, but in Winzip 7, the renaming is actually done automatically. You'll need only six files to run the programs (see list below):
designs.xl_ ---------> designs.xls
eed22.xl_ ---------> eed22.xla
er_test.xl_ ---------> er_test.xls
er22.xl_ ---------> er22.xla
erhelp.hl_ ---------> erhelp.hlp
Again, this is not a simple renaming, but extraction and renaming in one step. This should finally give you a running version of our software. If not, let us know!
8. ER and EED and Excel 2000
Based on our experience, the current versions of ER and EED should run in Excel 2000. One ought to make sure, however, that the Solver-Addin in properly installed (see 2).