Choosing values free of NA

[rbelhach95]

Dear Tom,

I have a vector series consisting of 400+ observation some of which are %na, I would like to choose 61 values among the 400+ free of %na and repeat generation of 61 values free of %na 1,000 times. Is there a way to do this in Rats? Please advise
Many thanks in advance
Rachid

[TomDoan]

Something like:

set source 1 450 = %if(%ranflip(.95),%ran(1.0),%na)
*
sample(smpl=%valid(source)) source / goodsource
boot entries 1 61 1 %nobs
set draw 1 61 = goodsource(entries(t))

You use SAMPLE with get rid of the NA's then use BOOT to sample the valid ones.

[rbelhach95]

Many thanks for your inputs. Does Rats 6.2 version contain the function %ranflip()? is %ranpermute(n) an equivalent function?
Thanks
Rachid

[TomDoan]

Many thanks for your inputs. Does Rats 6.2 version contain the function %ranflip()? is %ranpermute(n) an equivalent function?
Thanks
Rachid

No, but that's only to generate sample data. You have your own version of "source".

%RANFLIP(p) was added with 8.2. It can be done in earlier versions of RATS with (%uniform(0,1)<p).

[rbelhach95]

Many thanks Tom
Rachid

[rbelhach95]

Dear Tom,

Hope all is well. When I print / draw (code is below) I still get few NA among the 61 observations. Please advise
Many Thanks
Rachid

Code: Select all

do rep = 1,100
set v  = %ran(1.0)
...........
............
set R1(rep) = -psinustr0+psinustr1+estr
sample(smpl=%valid(R1(rep))) R1(rep) / goodsource
boot entries 1 61 1 %nobs
set draw 1 61 = goodsource(entries(t))
end do rep

[TomDoan]

put the 1 at the end of this:

sample(smpl=%valid(R1(rep))) R1(rep) / goodsource 1

Otherwise it might shift the data if R1 has missing values at the start.

[rbelhach95]

Many thanks Tom for the feedback. I have added 1 after goodsource, but I still get NA in draw. Please advise
Thanks
Rachid

Code: Select all

ENTRY         DRAW
 1971:03   0.009987129896
 1971:04   0.089264501125
 1971:05  -0.066277303726
 1971:06   0.034482091127
 1971:07        NA
 1971:08   0.028019166636
 1971:09        NA
 1971:10   0.002410742052
 1971:11   0.009987129896
 1971:12   0.011234568583
 1972:01   0.196553099441
 1972:02  -0.090971325404
 1972:03        NA
 1972:04   0.096994397199
 1972:05  -0.122666725915
 1972:06  -0.004744943820
 1972:07        NA
 1972:08   0.130441242877
 1972:09        NA
 1972:10   0.058699953278
 1972:11  -0.006844476663
 1972:12   0.051097005690
 1973:01   0.000913463102
 1973:02        NA
 1973:03  -0.019589709634
 1973:04   0.028831200672
 1973:05        NA
 1973:06   0.013102866851
 1973:07   0.040987096204
 1973:08   0.017601301523
 1973:09        NA
 1973:10  -0.067871752675
 1973:11   0.038154769307
 1973:12   0.075112074084
 1974:01   0.011234568583
 1974:02   0.055478393405
 1974:03   0.051117280703
 1974:04   0.028831200672
 1974:05  -0.015470479007
 1974:06   0.009606881975
 1974:07   0.050848591308
 1974:08  -0.088281064833
 1974:09   0.019939464135
 1974:10   0.031441730602
 1974:11   0.069240473027
 1974:12  -0.006844476663
 1975:01   0.009943722821
 1975:02   0.089264501125
 1975:03   0.089370631102
 1975:04   0.095017754493
 1975:05  -0.009238696336
 1975:06  -0.005077467003
 1975:07  -0.035146571003
 1975:08   0.052829071588
 1975:09   0.098658641980
 1975:10   0.020255125054
 1975:11   0.002410742052
 1975:12  -0.013834376261
 1976:01   0.073641693926
 1976:02  -0.004495060021

[TomDoan]

You'll have to post your whole program. It works fine as far as I can tell.

[rbelhach95]

Dear Tom,

Hope all is well. Kindly find the attached code, I still have the same problem of getting NAs. I also run to a problem when N=number of replication is over 10,000, the code runs for a while then it shows "RATS NOT RESPONDING" but for 5000 or less it works fine. Please advise
Thanks
Rachid

[TomDoan]

Why is the SAMPLE outside the DO REP loop?

end do rep

sample(smpl=%valid(R1(rep))) R1(rep) / goodsource 1
boot entries 1 61 1 %nobs
set draw 1 61 = goodsource(entries(t))
print / R1

[rbelhach95]

Thanks Tom for the feedback. Should the end of the code look like this:

Code: Select all

set psinustr1 = log(m_1)
set R1(rep) = -psinustr0+psinustr1+estr
sample(smpl=%valid(R1(rep))) R1(rep) / goodsource 1
boot entries 1 61 1 %nobs
set draw 1 61 = goodsource(entries(t))
end do rep
print / draw

Many thanks
Rachid

[TomDoan]

While that's better than using REP outside the REP loop, I don't see what you're trying to accomplish. This will generate a new DRAW series each time through the REP loop and then overwrite it the next time through. It would help you to learn how to use "pseudo-code" to organize what you're doing first without being as concerned about the precise programming code:

...written in a pseudo-code that resembles a computer language but
avoids excessive detail by lapsing from time to time into ordinary
English. This lets us specify quite a bit of the program before we
have worked out all aspects of it. On larger programs, it is
valuable to begin with pseudo-code and refine it in stages until
it is all executable. This way you can revise and improve the
design at a high level without writing any executable code, yet
remain close to a form which can be made executable.
(Software Tools, Kernighan, Plauger)

Kernighan is one of the co-authors of the C programming language (among many other things), so he knows a thing or two about how to program.

Also, please get in the habit of indenting your loops.

[rbelhach95]

Many thanks for your advice
Rachid