"Insufficient number of observations"- problem

[lomfy]

Hi there!
I try to run regression for 96 sets of dependent variables ("rs") and independent variables (mam) of the following simple form...

for !i=1 to 96
equation eq!i.ls rs!i=c(1)+c(2)*mam!i
next

which should result in the same equation objects as...

equation eq1.ls(h) rs1=c(1)+c(2)*mam1
....
equation eq8.ls(h) rs8=c(1)+c(2)*mam8
equation eq9.ls(h) rs9=c(1)+c(2)*mam9
equation eq10.ls(h) rs10=c(1)+c(2)*mam10
equation eq11.ls(h) rs11=c(1)+c(2)*mam11
equation eq12.ls(h) rs12=c(1)+c(2)*mam12
...

However, with some of the datasets there appears to be an issue as i get the message "insufficent number of observations" - no equations are written for these datasets and the process terminates. I've checked the datasets thoroughly, they are all of the same lenght (241 entries).

Any hint how to automatically get estimations for all datasets? (The mistake also appears if I create the equations manually, as indicated above [without the "for" clause])

Thanks a lot!!

Wolfgang

[startz]

rs13 is all NA and mam13 is all zero

[lomfy]

Hi startz!

Yes, I'm aware of that ... in the first place I thought this might e the only problem.
However, even if I enter

for !i=14 to 96
equation eq!i.ls rs!i=c(1)+c(2)*mam!i
next

the program finds this mistakes, and even in datasets, that do not have this issue (see e.g. the screenshot example with rs28 and mam28m, which are perfectly ok as far a I can see!). It's about those instances that I care...I'll ignore the empty data sets (I think 13 is the only set with this problem with this issue in the whole database)...

What about the mistake with regards to those sets that are/look perfectly fine?

Thanks!

[lomfy]

Moreover:

When i run the programme repeatedly, the mistake occurs in different runs with different data sets!

For example:

- 1. time: when i ran it for 90 - 96, first there was a mistake at 92
- 2. time: when i ran it again (without change), there was a mistake at 90
- 3. time: when i ran it the third time, there was no mistake at all!

Each time there is a mistake and i doubleclick the respective equation it says "Equation estimates are not valid".

What can i do? If I had to estimate only those 96 equations I would simply write down the results as soon as I get them
(repeat the procedure until I collected reasonable results for all equations [without number 13])
-
the problem is: I have to estimate approx. 12.000 equations in this manner (120 times databases with 96 time series),
so I really need to automate this thing...!

Any help with this issue?

Thanks!

[startz]

There is certainly something very odd going on. Why don't you:
(1) Update your copy of of EViews.
(2) Post the workfile and program that actually cause the error.
(3) Hope for a GarethRescue.

[EViews Gareth]

I ran the following:

Code: Select all

for !i=1 to 96 if !i<>13 then equation eq!i.ls rs!i=c(1)+c(2)*mam!i endif next

(i.e. take your code and skip over the one problematic case that Startz pointed out), and ran it 50 times in a row. Every time it was able to estimate every equation fine.

[lomfy]

well....i'll get eviews 7 in about a week from now - then i'll try again! hopefully it's a problem with this version...i'let you know!

[EViews Gareth]

The screenshot you posted was from EViews 7...

[lomfy]

i wonder, as i ran the programme on my colleagues computer and the first thing she told me was that she did not have the latest eviews version
(so i didn't look up the version on her computer and assumed it was 6)...well if the mistake happens even on 7 that's not good :/ i'll have to wait and see whether my version will handle it...