Question
shahlaebrahimi opened this issue · 3 comments
shahlaebrahimi commented
@osofr Hi
I would greatly appreciate if you could let me know whether your code is suitable for my data set which is as follows. In fact, it is an unbalanced longitudinal data set with time varying features. I want to predict RE to CS using x1 to x5 features.
ID year RE to CS x1 x2 x3 x4 x5
1 1 0.06039 1.28102412 0.022933584 0.87453816 1.216366609 0.06094049
1 2 0.01064 1.270012471 0.00645422 0.820672937 1.004861122 -0.014079609
1 3 -0.45597 1.052890304 -0.059378881 0.922421512 0.729264145 0.020475912
1 4 -0.32539 1.113115232 -0.01522879 0.858878436 0.809737564 0.07603735
1 5 -0.56657 1.219644234 -0.058675441 0.887087711 0.484342194 0.009777888
2 1 1.25097 1.06226374 0.107020836 0.814602294 0.835928139 0.19996023
2 2 1.35725 1.055785531 0.081916221 0.879486383 0.686727862 0.142627013
2 3 0.00719 0.970588058 0.076063501 0.906774596 0.809795658 0.165915285
2 4 1.20019 1.058995743 0.130202682 0.818111675 0.875989179 0.23445163
2 5 2.23481 1.12452475 0.147841049 0.758709609 1.079924775 0.276444488
2 6 1.34048 1.599780804 0.262461269 0.546150712 1.312740749 0.369478637
2 7 2.04740 1.575608388 0.262096474 0.564481097 1.156476191 0.3486243
2 8 2.34589 1.544272968 0.240910847 0.590728825 1.076969981 0.325612011
2 9 2.24994 1.721707641 0.215246493 0.552290866 0.841010871 0.293499528
2 10 2.28261 1.723163256 0.208630134 0.533981319 0.786512171 0.293033271
2 11 1.79821 1.630677468 0.186234679 0.547718673 0.728193067 0.273576931
2 12 2.82772 2.17231306 0.319454809 0.441392998 0.94698478 0.427395498
3 1 -0.30317 0.874395008 -0.034676249 0.79350188 0.609515013 -0.002631637
3 2 -1.65989 0.825239215 -0.14194334 0.952212806 0.572879612 -0.019154984
``
`Best regards,
osofr commented
Hi,
gridisl relies on machine learning algorithms implemented in xbgboost and
h2o R packages. The only thing gridisl does is making it easier to use
those two packages with a unified syntax. In short, the answer is yes, you
can use gridisl to do prediction for your longitudinal data. I recommend
looking into the aforementioned two packages for learning about specific
prediction algorithms to try on your data. Also please see examples in
tests folder of gridisl for examples of various algorithms that can be used.
Best,
Oleg
~~ sent from mobile device ~~
On Mar 6, 2017 02:10, "shahlaebrahimi" <notifications@github.com> wrote:
Hi
I would greatly appreciate if you could let me know whether your code is
suitable for my data set which is as follows. In fact, it is an unbalanced
longitudinal data set with time varying features. I want to predict RE to
CS using x1 to x5 features.
ID year RE to CS x1 x2 x3 x4 x5
1 1 0.06039 1.28102412 0.022933584 0.87453816 1.216366609 0.06094049
1 2 0.01064 1.270012471 0.00645422 0.820672937 1.004861122 -0.014079609
1 3 -0.45597 1.052890304 -0.059378881 0.922421512 0.729264145 0.020475912
1 4 -0.32539 1.113115232 -0.01522879 0.858878436 0.809737564 0.07603735
1 5 -0.56657 1.219644234 -0.058675441 0.887087711 0.484342194 0.009777888
2 1 1.25097 1.06226374 0.107020836 0.814602294 0.835928139 0.19996023
2 2 1.35725 1.055785531 0.081916221 0.879486383 0.686727862 0.142627013
2 3 0.00719 0.970588058 0.076063501 0.906774596 0.809795658 0.165915285
2 4 1.20019 1.058995743 0.130202682 0.818111675 0.875989179 0.23445163
2 5 2.23481 1.12452475 0.147841049 0.758709609 1.079924775 0.276444488
2 6 1.34048 1.599780804 0.262461269 0.546150712 1.312740749 0.369478637
2 7 2.04740 1.575608388 0.262096474 0.564481097 1.156476191 0.3486243
2 8 2.34589 1.544272968 0.240910847 0.590728825 1.076969981 0.325612011
2 9 2.24994 1.721707641 0.215246493 0.552290866 0.841010871 0.293499528
2 10 2.28261 1.723163256 0.208630134 0.533981319 0.786512171 0.293033271
2 11 1.79821 1.630677468 0.186234679 0.547718673 0.728193067 0.273576931
2 12 2.82772 2.17231306 0.319454809 0.441392998 0.94698478 0.427395498
3 1 -0.30317 0.874395008 -0.034676249 0.79350188 0.609515013 -0.002631637
3 2 -1.65989 0.825239215 -0.14194334 0.952212806 0.572879612 -0.019154984
``
`Best regards,
—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub
<#1>, or mute the thread
<https://github.com/notifications/unsubscribe-auth/AGaPTo8Wj-2--hpCAA1Vwkey66GBL2A6ks5ri9ufgaJpZM4MT8xv>
.
shahlaebrahimi commented
@osofr Hi
I thank you very much for your time and consideration. In fact, since you ignored the time varying features like "wtkg", "bmi", and "haz", I wans not sure if it could deal with time varying features among different subjects or not.
Best regards,
osofr commented
These time varying features are heavily correlated with the outcomes and
that's why they were excluded. Time variable is a time varying feature, for
example, and it was used for modeling. Nothing prevents you from including
any time varying feature you like in you model.
~~ sent from mobile device ~~
…On Mar 7, 2017 08:40, "shahlaebrahimi" ***@***.***> wrote:
@osofr <https://github.com/osofr> Hi
I thank you very much for your time and consideration. In fact, since you
ignored the time varying features like "wtkg", "bmi", and "haz", I wans not
sure if it could deal with time varying features among different subjects
or not.
Best regards,
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#1 (comment)>, or mute
the thread
<https://github.com/notifications/unsubscribe-auth/AGaPTtjWmwGB19f06Itt3wHhL8H2aVWCks5rjYhogaJpZM4MT8xv>
.