How to use .predict() method

[Jack90_]

I am trying to apply a forecasted model with OLS regression to a test dataset, but when I use the .predict() method, I have this error:

ValueError: shapes (2938,4) and (1583,5) not aligned: 4 (dim 1) != 1583 (dim 0)

I add this replicable example of the problem where I obtain the same error (with different shapes because dataset is different):

from sklearn.datasets import load_iris
from sklearn.model_selection import train_test_split
import statsmodels.api as sm
iris_data = load_iris()
iris_dataframe = pd.DataFrame(iris_data.data, columns=iris_data.feature_names)

X_train, X_test = train_test_split(iris_dataframe, test_size=0.2, random_state=1)

X_train.head()
X_test.head()

model_test = sm.OLS(X_train["petal width (cm)"],X_train.iloc[:,0:3])
results_test=model_test.fit()
print(results_test.summary())

test_preds=model_test.predict(X_test)

As you can see, when I try to create test_preds, I have the sample error. I am using wrongly the .predict method? I am using as input the dataset X_test, that has the same columns number and names of the training dataset X_train.

[Larz60+]

Is this related?

[Jack90_]

Is this related?

I think no. That question uses the Iris dataframe as an example like I did, but it's not on .predict() method.

[Jack90_]

I found the solution. The object to be passed to .predict() is not the model object, but the fitobject. In this way it works:

test_preds=results_test.predict(X_test)

[Larz60+]

Thanks for sharing solution