home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!paladin.american.edu!auvm!UCRAC1.UCR.EDU!JAMES
- Message-ID: <930125101811.2041112d@UCRAC1.UCR.EDU>
- Newsgroups: bit.listserv.stat-l
- Date: Mon, 25 Jan 1993 10:18:11 -0800
- Sender: STATISTICAL CONSULTING <STAT-L@MCGILL1.BITNET>
- From: JAMES@UCRAC1.UCR.EDU
- Subject: excel: an example of how NOT to send/apply zaps
- Lines: 100
-
- This somehow makes me think of that banal saying that 'When the only tool you
- have is a hammer, all problems look like nails' (or something like that)...
- Anyway folks who insist on using spreadsheets for statistical analyses might
- possibly be interested in this.
-
- What a nightmare it would be if fixes/zaps/patches to major statiscal packages
- were distributed in similar fashion to that apparently used by Excel!
-
- james ssemakula
- u.c. riverside
-
- ------------------- begin forwarded message-----------------------------------
- From: SMTP%"BSCS-L%EMUVM1.BitNet@pucc.PRINCETON.EDU" 25-JAN-1993 00:32:31.77
- To: JAMES
- CC:
- Subj: EXCEL 4.0 REGRESSION BUG AND FIX
-
- Date: Sun, 24 Jan 1993 21:56:00 EST
- Reply-To: Business School Computing Support
- <BSCS-L%EMUVM1.BitNet@pucc.PRINCETON.EDU>
- Sender: Business School Computing Support
- <BSCS-L%EMUVM1.BitNet@pucc.PRINCETON.EDU>
- From: "Eric R. Fliegel" <EFLIEGEL%EMUBUS.BitNet@pucc.PRINCETON.EDU>
- Subject: EXCEL 4.0 REGRESSION BUG AND FIX
- To: Multiple recipients of list BSCS-L
- <BSCS-L%EMUVM1.BitNet@pucc.PRINCETON.EDU>
-
- >>> I THOUGHT SOME OF YOU WOULD BE INTERESTED IN THIS.
- HERE IS THE MESSAGE THAT MY EXCEL CONTACT SENT TO HIS TECHNICAL GROUP:
-
- Eric Fliegel at Emory University is reporting the following two problems in
- Excel's regression function:
-
- 1. It is possible for an independent variable to be dependent on
- another independent variable without Excel generating an error message.
- It seems Excel should not allow a user to do this.
-
- 2. A regression is run on the following data:
-
- X - independent variable Y - dependent variable
- ===================== ====================
- .6 220
- .62 200
- .58 280
- .6 250
- .64 190
- .62 240
- .
-
- When Excel runs this regression, Excel reports the following values:
-
- Significant F = .0301
- P Value = .0216
-
- According to Emory, the Significant F and P Values should be the same.
- SAS reports the values as being the same on the identical data set.
-
- >>> HERE WAS THEIR RESPONSE:
-
- We have some input on one of the problems you reported in Excel's
- regression function. Our description of the problem as it appears in
- our database is listed below. Also listed are changes that can be made
- to the Excel Add-in Macro REGRESS.XLA which will correct the problem
- (REGRESS.XLA) is the macro used to perform the regression in Excel. We
- plan to implement the fix below in the next version of Excel.
-
- Database entry is as follows:
-
- The P-value output from the ATP Regression procedure is incorrect.
- This problem was tracked down by one of our customers to cells B383 and
- B391 in REGRESS.XLA. Changing the 12,1,1,1 strings in the OFFSET()
- formulas in these two cells to 11,1,1,1 should correct the problem with
- the P-value and also cause the P-value to match the Significance F value.
-
- Line 383 is:
- =FORMULA(OUTPUT(TDIST(ABS(OFFSET(TEXTREF(soutrng_txt),16,3,1,1)),OFFSET
- (TEXTREF(soutrng_txt),12,1,1,1),2)),OFFSET(TEXTREF(soutrng_txt),16,4,1,1))
-
- Should be:
- =FORMULA(OUTPUT(TDIST(ABS(OFFSET(TEXTREF(soutrng_txt),16,3,1,1)),OFFSET
- (TEXTREF(soutrng_txt),11,1,1,1),2)),OFFSET(TEXTREF(soutrng_txt),16,4,1,1))
-
- And line 391 is:
- =FORMULA(OUTPUT(TDIST(ABS(OFFSET(TEXTREF(soutrng_txt),regress_rows1+
- Counter,3,1,1)),OFFSET(TEXTREF(soutrng_txt),12,1,1,1),2)),OFFSET(TEXTREF(
- soutrng_txt),regress_rows1+Counter,4,1,1))
-
- And should be:
- =FORMULA(OUTPUT(TDIST(ABS(OFFSET(TEXTREF(soutrng_txt),regress_rows1+
- Counter,3,1,1)),OFFSET(TEXTREF(soutrng_txt),11,1,1,1),2)),OFFSET(TEXTREF(
- soutrng_txt),regress_rows1+Counter,4,1,1))
-
- >>> FOR YOUR INFORMATION, TO EDIT AN ADD-IN (.XLA) FILE, HOLD THE <SHIFT> KEY
- DOWN WHEN YOU SELECT THE FILE. THE REGRESS.XLA FILE IS IN THE
- \EXCEL4\LIBRARY\ANALYSIS SUBDIRECTORY. SIMPLY MAKE THE CHANGES TO LINES 383
- AND 391 AND IT WORKS! FOR MACINTOSH USERS, THE FILE REGRESSION IS IN THE
- ANALYSIS TOOLS FOLDER WHICH IS IN THE MACRO LIBRARY FOLDER WHICH IS IN THE
- EXCEL4 FOLDER. NOTE THAT THE LINES TO MODIFY ARE 384 AND 392. HOPE THIS
- HELPS. ERic Fliegel
- ------------------------end forwarded message----------------------------
-