concat() got an unexpected keyword argument 'join_axes'

Question

I am trying to use pymc3 in an ipynb on Google Colab.

Here is my code:

regression_conjugate = pm.Model()
with regression_conjugate:
  sigma2 = pm.InverseGamma("sigma2",alpha = 0.5*nu0,beta=0.5*lam0)
  sigma = pm.math.sqrt(sigma2)
  a = pm.Normal("a",mu = b0[0],sd = sigma*sd0[0])
  b = pm.Normal("b",mu = b0[1],sd = sigma*sd0[1])
  y_hat = a+b*x
  likelihood = pm.Normal("y",mu = y_hat,sd = sigma,observed = y)

n_draws = 50
n_chains = 4
n_tune = 1000
with regression_conjugate:
  trace = pm.sample(draws = n_draws, chains=n_chains,tune=n_tune,random_seed=123)

print(pm.summary(trace))

However, this outputs the following:

TypeError                                 Traceback (most recent call last)
<ipython-input-44-4e1a1fef1a74> in <module>()
     26   trace = pm.sample(draws = n_draws, chains=n_chains,tune=n_tune,random_seed=123)
     27 
---> 28 print(pm.summary(trace))

TypeError: concat() got an unexpected keyword argument 'join_axes'

please let me know if anyone who understands.

pymc3 : 3.7

pandas : 1.0.3

I tried to run PyMC3 but got an `Compiler Deprecated error`. Turns out you might be using newer version of Python3 than Theano compiler which needs to be configured for PyMC3 can support. Read [this](http://www.deeplearning.net/software/theano/requirements.html). Try to edit your post with more information and code that can be run. You are also missing import statements. — sanitizedUser, Apr 22 '20 at 13:14
What happens if you just do `pm.summary(trace)` without the `print`? — merv, Apr 22 '20 at 18:23
I'm sorry for the lack of information. I tried to upgrade version of pymc3 to 3.8 which is the latest version. So this problem has been solved. Thank you so much! — 渡邊彰久, Apr 23 '20 at 06:47

score 5 · Answer 1 · answered Oct 02 '20 at 17:08

"join_axes" was deprecated in version 0.25 for some reason. You can achieve the same effect by reindexing.

#won't work:
df3 = pd.concat([df1, df2], axis=1,join_axes=[df1.index]) #won't work

#instead:
df3 = pd.concat([df1, df2], axis=1)
df3 = df3.reindex(df1.index)

score 1 · Answer 2 · answered Aug 16 '21 at 14:54

On the other hand, if you came here because you tried to used join_axes with columns, try this instead:

# the columns of df3 and only the columns of df4 that are common
df_new = pd.concat([df3, df4], axis=1)
df_new[df3.columns]

With df3 as

    C   D   E
 0  C1  D1  E1
 1  C2  D2  E2
 2  C3  D3  E3

and df4 as

    D   E   F
 0  D5  E5  F5
 1  D6  E6  F6
 2  D7  E7  F7

the result that will be produced is

    C   D   D   E   E
 0  C1  D1  D5  E1  E5
 1  C2  D2  D6  E2  E6
 2  C3  D3  D7  E3  E7

score 0 · Answer 3 · answered Jun 10 '20 at 09:21

0

As 渡邊彰久 said in the comments upgrading PyMC3 to version 3.8 solved the problem.

answered Jun 10 '20 at 09:21

Giovanni Fabbretti

81
8

score 0 · Answer 4 · answered Jul 01 '21 at 15:44

0

The "join_axes" function is deprecated. As a result of my research, I came to the following;

instead of:

pd.concat ([df1, df2], join_axes = [df1.columns])

You can try this:

pd.concat([df1, df2.reindex(columns = df1.columns)], ignore_index = True)

answered Jul 01 '21 at 15:44

Ertugrul Kuyucu

1

concat() got an unexpected keyword argument 'join_axes'

4 Answers4