Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Compatibility issue with statsmodels 0.18.0 #897

Closed
wjsi opened this issue Oct 3, 2021 · 8 comments · Fixed by Kaggle/docker-python#1122
Closed

Compatibility issue with statsmodels 0.18.0 #897

wjsi opened this issue Oct 3, 2021 · 8 comments · Fixed by Kaggle/docker-python#1122
Labels

Comments

@wjsi
Copy link

wjsi commented Oct 3, 2021

The problem:

Since v0.18.0, statsmodels stops supporting unbiased and uses adjusted (statsmodels/statsmodels@3d6c388#diff-771ed9ae3d316dec0997427f2182d015dbfdaf6935335428bbf177ce1968bc72), and use of statsmodels.tsa.stattools.acf in tsfresh.feature_extraction.ceature_calculators now causes TypeError. We need to use adjusted instead of unbiased in calls for acf in statsmodels later than v0.14.0.

@fkiraly
Copy link

fkiraly commented Oct 3, 2021

you mean 0.13.0, right? The update on Friday that also broke sktime since it depends on tsfresh?

See sktime/sktime#1478

We just verified that indeed reverting to 0.12.1 seems to resolve the issue (temporarily).

The fix in tsfresh would be to change the name of the argument.

@SirJohnFranklin
Copy link

SirJohnFranklin commented Oct 14, 2021

For your information, this has been changed too:
NotImplementedError: AR has been removed from statsmodels and replaced with statsmodels.tsa.ar_model.AutoReg

Maybe something like this:

from statsmodels.tsa.ar_model import AutoReg, ar_select_order
lags = ar_select_order(x_as_list, maxlag=k)
calculated_AR = AutoReg(x_as_list, lags=lags)
calculated_ar_params[k] = calculated_AR.fit(cov_type="robust").params

@queyrusi
Copy link

For your information, this has been changed too: NotImplementedError: AR has been removed from statsmodels and replaced with statsmodels.tsa.ar_model.AutoReg

Maybe something like this:

from statsmodels.tsa.ar_model import AutoReg, ar_select_order
lags = ar_select_order(x_as_list, maxlag=k)
calculated_AR = AutoReg(x_as_list, lags=lags)
calculated_ar_params[k] = calculated_AR.fit(cov_type="robust").params

Should just make sure lags refers to the .ar_lags attribute of ar_select_order(x_as_list, maxlag=k).
Works fine for me.

@theoohoho
Copy link

Hi ! Are there any updates?

@wlmszdnb
Copy link

i see this bug today. Though I turn statemodel to 0.12.1, but does't work? Help,pls.

@MassEast
Copy link

MassEast commented Nov 4, 2021

On 0.11 everything works fine.

@VGODIE
Copy link

VGODIE commented Dec 8, 2021

I am really looking forward to this update ! Are there any changes happened?

@kempa-liehr
Copy link
Collaborator

@SirJohnFranklin, thanks for the short demo of the new interface.

rosbo added a commit to Kaggle/docker-python that referenced this issue Jan 18, 2022
Remove unnecessary pin on statsmodels following tsfresh new release
including this fix: blue-yonder/tsfresh#897

http://b/215027867
rosbo added a commit to Kaggle/docker-python that referenced this issue Jan 19, 2022
* Upgrade base image to m88.

Remove unnecessary pin on statsmodels following tsfresh new release
including this fix: blue-yonder/tsfresh#897

http://b/215027867

* pin setuptools to 59.5.0 in torch build
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
9 participants