To all users of Amelia,
Using basis functions in case of time series cross section data seems to
improve amelia imputations substancially (at least in my datacase).
**In Amelia manual on define polynomials used as b0 + b1* t + b1* t^2 +
b1*t^3. I understand it is a typographic error and coefficients are
different (b0, b1, b2, b3..)
Spline function looks like improves slightly imputations (so far
splinetime=4) and it´s generally preferred for interpolation (wikipedia).
I`m just testing.
But LOESS or local regression, where on make a different regression for
each element based on the k-neighbors, using least squares regression over
a polynomial is really appealing. The tecnique is not widely used (I guess)
because until a few years ago computers didn´t allow for it.
James Honaker/Gary King achieve better imputations in "What to Do About
Missing Values in TSCS Data" with LOESS.
Meanwhile, and as far as I keep testing I have also changed from limit the
draws (, emburn=c(20,100)) to increase tolerance (,tolerance=0,001) and
results should now be more solids.
** Also trying to subset overimpute (a.out, subset=country=="france",
"Vtas") gives me this error:
Error in `[<-`(`*tmp*`, , j - 1, value = c(NA, NA, NA, NA, NA, NA, NA, :
subscript out of bounds
I think it could be for having variables (defined as "idvars" at
imputation but not at overimpute instruction) with "NA".
Someone has experience in changing basis functions or could give some help?
Jesus Borruel
PHD student
San Pablo-CEU University
Spain(EU)
Hi,
I’m using Amelia (version 1.7.2) in R (version 3.1.1) to impute missing values for my dataset which includes variables with known logical bounds. The problem is that the bounds argument doesn’t seem to work in the user front or in R and the program doesn’t send out an error message. I’ve tried following your example with the free trade dataset and the bounds don’t hold there either.
Any ideas what is going on? Any advice you could offer would really be appreciated.
Thanks,
Kristin
Sir,
I use try and error method to improve imputations with a database of firm's
variables (time series cross section data with collinearity:1800
rows-observations, 163 firms, 43 column-variables,) .
It has been helpful to limit number of draws to 100 (,emburn=c(20,100))
after getting convergence with 400 and 500 draws and seen disperse grafics
(a.out, dims=1,m=5) visually converge at draws 70-80.
Also taking out variables for collinearity, adding lags and leads and using
polynomials (polytime =3, much better than polytime =2 or not polytime at
all).
My problem is I'm not able to know how use LOESS smoothing to create basis
functions. From "What to do about missing values in TSCS data":*
(Finally,we also ran a third set of 120 imputation models,this time using
LOESS smoothing to create the basis functions... LOESS based smoothing
provides a clear advantage over polynomial smoothing: almost as many points
are captured by the 90% confidence intervals as for the polynomials, but
the LOESS-based intervals are narrower in almost all cases, especially when
the polynomial-based intervals are largest.)*
I´ve read carefully Amelia II, What to do with TSCS data and also posting
lists in Amelia archives but got no conclusion.
I'd appreciate any help any hint
Jesus Borruel
PH student
San Pablo-CEU University
Hi Alicia & Matt
Many thanks for your exhaustive answers. I'm very grateful for your kind help.
Le Jeudi 18 septembre 2014 14h51, Alicia Doyle Lynch <aliciadlynch(a)gmail.com> a écrit :
Hi Mouna -
I wouldn't necessarily agree that the "best result" is the mean pre-imputation value for each variable. Part of the reason for doing imputation is because attrition, non-response, etc. may have introduced biases to the sample. As such, the imputation process creates a variety of estimates that will vary around that pre-imputation mean. When you later conduct your analyses on the imputed data, the analyses will account for this variation - it's an important piece of information.
To compare your pre-imputation mean to post-imputation means is important - post-imputation values that are very different from pre-imputation values suggest you may need to rethink your imputation model - but as long as there are no "red flags" - I would suggest letting the imputation results stand as they are.
-Alicia
On Thu, Sep 18, 2014 at 8:09 AM, mouna kessentini <kessentini_mouna(a)yahoo.fr> wrote:
Dear professors,
>
>
>I have performed multiple imputation for panel data with Amelia II. After 10 iterations, I noticed that the imputed data concerning for example my variable « X1 » is better after the imputation number 4. However, the imputed data concerning the variable « X2 » is better after the imputation number 2.
>Is it possible to compare the imputed data for every variable in each iteration and retain the best result knowing that the comparison criteria for the best result is the mean value for each variable before and after imputation? Thank you in advance for your answer.
>
>
>
>
>
>Best regards,
>Mouna kessentini
>
>
>
>--
>Amelia mailing list served by HUIT
>[Un]Subscribe/View Archive: http://lists.gking.harvard.edu/?info=amelia
>More info about Amelia: http://gking.harvard.edu/amelia
>Amelia mailing list
>Amelia(a)lists.gking.harvard.edu
>
>To unsubscribe from this list or get other information:
>
>https://lists.gking.harvard.edu/mailman/listinfo/amelia
>
--
Alicia Doyle Lynch, Ph.D.
Boston College, Lynch School of Education
Department of Developmental and Educational Psychology
Chestnut Hill, MA 02467
Phone: (617) 552-6437
e-mail: doylead(a)bc.edu
Best regards
Mouna
Dear professors,
I have performed multiple imputation for panel data with Amelia II. After 10 iterations, I noticed that the imputed data concerning for example my variable « X1 » is better after the imputation number 4. However, the imputed data concerning the variable « X2 » is better after the imputation number 2.
Is it possible to compare the imputed data for every variable in each iteration and retain the best result knowing that the comparison criteria for the best result is the mean value for each variable before and after imputation? Thank you in advance for your answer.
Best regards,
Mouna kessentini