如何用列迭代填充Pandas Dataframe

Question

I am trying to create a pandas dataframe with iteratively counting statisitcs from another dataframe, it goes through columns (that filtered with regex). How could i create a result Dataframe? Input dataframe:

    In [4]: control.head()
    Out[4]:
  Patient Gender  Age  Left-Lateral-Ventricle_NVoxels  Left-Inf-Lat- 
Vent_NVoxels  ...  supramarginal_CurvInd_lh
0    P008      M   30                            9414                        
311  ...                       7.5
1    P013      F   35                            7668                         
85  ...                      10.4
2    P018      F   27                            7350                        
202  ...                       8.0
3    P033      F   55                            7548                        
372  ...                       9.2
4    P036      F   31                            8598                         
48  ...                       8.0

    [5 rows x 930 columns]

I wrote a code to count statistic, but stuck on creating result pandas dataframe

def select_volumes(group_c,group_k):
    Select_list = ["Amygdala", "Hippocampus", "Lateral-Ventricle", 
"Pallidum", "Putamen", "Thalamus"]
    Side = ["Left", "Right"]
    for s in Side:
        for struct in Select_list:
            volumes_c = group_c.filter(regex="^(?=.*"+s+")(?=.*"+struct+") 
   (?=.*Volume)")
            volumes_k = group_k.filter(regex="^(?=.*"+s+")(?=.*"+struct+") 
   (?=.*Volume)")
            k = cohens_d(volumes_c, volumes_k)
            meand = volumes_c.mean()
            result_df = pd.Dataframe(
{
     "Cohen's norm": some result
     "Mean Value": meand
}
)
            return k

function select_volumes gives me the result:

Left-Amygdala_Volume_mm3   -0.29729
dtype: float64
Left-Hippocampus_Volume_mm3    0.33139
dtype: float64
Left-Lateral-Ventricle_Volume_mm3   -0.111853
dtype: float64
Left-Pallidum_Volume_mm3    0.28857
dtype: float64
Left-Putamen_Volume_mm3    0.696645
dtype: float64
Left-Thalamus-Proper_Volume_mm3    0.772492
dtype: float64
Right-Amygdala_Volume_mm3   -0.358333
dtype: float64
Right-Hippocampus_Volume_mm3    0.275668
dtype: float64
Right-Lateral-Ventricle_Volume_mm3   -0.092283
dtype: float64
Right-Pallidum_Volume_mm3    0.279258
dtype: float64
Right-Putamen_Volume_mm3    0.484879
dtype: float64
Right-Thalamus-Proper_Volume_mm3    0.809775
dtype: float64

I want Left-Amygdala_Volume_mm3 ... be the row with value -0.29729 with column name Cohen's d be the column for every Select_list: example, how dataframe should looks

Answer 1

I still cannot really understand how and where, but you showed that somewhere in the function you were able to build a float64 Series containing for example Left-Amygdala_Volume_mm3 as index and -0.29729 as value. And I assume that at the same time, you have the value of meand for the same index value.

More exactly I will assume:

k = pd.Series([-0.29729], dtype=np.float64,index=['Left-Amygdala_Volume_mm3'])

because it prints as:

print(k)

Left-Amygdala_Volume_mm3   -0.29729
dtype: float64

At the same time, I assume that meand is also a similar Series. So we will access its value as meand.iloc[0] (lets say value is 9174.1)

You should combine them to build the content of a row:

row = k.reset_index().iloc[0].tolist() + [meand.iloc[0]]

In the example we have row : ['Left-Amygdala_Volume_mm3', -0.29729, 9174.1]

So you now need to build a large list of that rows:

def select_volumes(group_c,group_k):
    Select_list = ["Amygdala", "Hippocampus", "Lateral-Ventricle", 
"Pallidum", "Putamen", "Thalamus"]
    Side = ["Left", "Right"]
    data = []
    for s in Side:
        for struct in Select_list:
            volumes_c = group_c.filter(regex="^(?=.*"+s+")(?=.*"+struct+") 
   (?=.*Volume)")
            volumes_k = group_k.filter(regex="^(?=.*"+s+")(?=.*"+struct+") 
   (?=.*Volume)")
            k = cohens_d(volumes_c, volumes_k)
            meand = volumes_c.mean()

            # build a row of result df
            data.append(k.reset_index().iloc[0].tolist() + [meand.iloc[0]])

    # after the loop combine the rows into a dataframe and return it:
    result = pd.DataFrame(data, columns=['index', "Cohen's d", 'Mean']).set_index('index')
    return result

Answer 2

I write to pd.Dataframe inside a function:

k = cohens_d(volumes_c, volumes_k)
meand = volumes_c.mean()    
volumes_df.append([cohen.index[0],cohen.values[0], meand)
return volumes_df

and out of a function I call pd.Dataframe with:

finaldf=pd.DataFrame(select_volumes(control,patolog))
finaldf.columns=['Structure','Cohensd','Meand')