Pandas DataFrame - 'cannot astype a datetimelike from [datetime64[ns]] to [float64]' when using ols/linear regression

Question

I have a DataFrame as follows:

	Ticker	Date	Close
0	ADBE	2016-02-16	78.88
1	ADBE	2016-02-17	81.85
2	ADBE	2016-02-18	80.53
3	ADBE	2016-02-19	80.87
4	ADBE	2016-02-22	83.80
5	ADBE	2016-02-23	83.07

...and so on. The Date column is the issue. I'm trying to get the linear regression of the Date column with the Close column:

ols1 = pd.ols(y=ADBE['Close'], x=ADBE['Date'], intercept=True)

I get the following error:

TypeError: cannot astype a datetimelike from [datetime64[ns]] to [float64]

I've tried multiple ways of getting rid of this error, for examples:

dates_input = ADBE['Date'].values.astype('datetime64[D]')

dates_input = ADBE['Date'].values.astype('float')

The second dates_input attempt returns the type as pandas.core.series.Series but I still get an error message.

Does anyone know how to get the Date column to work and get rid of this TypeError?

Chinch

Answer 1

You need:

ADBE['Date'] = ADBE['Date'].values.astype(float)

and then:

ols1 = pd.ols(y=ADBE['Close'], x=ADBE['Date'], intercept=True)

Avidity

I made a mistake; for others who might make the same mistake; be careful not to use Series.astype(float) (as in; pd.Series(ADBE['Date'].values).astype(float); notice @jezrael's answer uses ndarray.astype(float) (directly on the .values.astype(float)) – Wineshop

I was getting the error TypeError: cannot astype a datetimelike from [datetime64[ns]] to [int32] when trying to run the line df_['Date']=(df['Date'].astype(int)// 109).astype('U10') Changing above line to df['Date']=(df['Date'].values.astype(int)// 109).astype('U10') solved by issue. – Azevedo