Running coint_johansen cointegration test gives : LinAlgError: Matrix is not positive definite

Question

I am pretty new to mulltivariate time series, I am trying to make a VAR model with 108 predictors and 1 target variable. While performing the Johansen Cointegration Test, I am getting an error

LinAlgError: Matrix is not positive definite

My code is :

'''python
def cointegration_test(df, alpha=0.05): 
    """Perform Johanson's Cointegration Test and Report Summary"""
    out = coint_johansen(df,-1,5)
    d = {'0.90':0, '0.95':1, '0.99':2}
    traces = out.lr1
    cvts = out.cvt[:, d[str(1-alpha)]]
    def adjust(val, length= 6): return str(val).ljust(length)

    # Summary
    print('Name   ::  Test Stat > C(95%)    =>   Signif  \n', '--'*20)
    for col, trace, cvt in zip(df.columns, traces, cvts):
        print(adjust(col), ':: ', adjust(round(trace,2), 9), ">", adjust(cvt, 8), ' =>  ' , trace > cvt)

cointegration_test(g)
'''

• Where g is my Time Series Dataframe of shape (48 rows × 109 columns) . rows are the date-time index and columns are the predictors/variables.

• Data in few columns ranging from 0-1(For ex : Consumer Price Index) and others in the range of millions(For ex : Population, GDP).

• There are columns in the dataframe which contain negative terms as well(For eg : Change in Employment)

• Few columns also contain zero in them

But when i pass the dataframe after making all the column stationary using

g = g.diff().dropna().diff().dropna()

and then pass the differenced dataframe to cointegration_test its giving error as :

LinAlgError: Matrix is not positive definite

As far as my understanding goes Matrix is not positive definite means that the Eigen values associated with it are non-positive.And Eigen-values are only possible for square matrix, but given the data which I am feeding is a non-square...

How can I solve this problem? Where should I look next? Would appreciate any help.

Thanks