根据两个值选择 Pandas Dataframe 中的部分行

Question

I'm trying to select a subset of rows and columns from a pandas dataframe that I'm going to eventually graph. My data is currently structured:

                  0       2        3  ...      177     178  Timestamp
1                                     ...                            
6:54:36   7/26/2019   -35.0   -34.75  ...     8.75     9.0   06:54:36  
 500 a  7/26/2019  3880.0  4068.00  ...  4562.00  4398.0   06:54:36
 500 b  7/26/2019  3462.0  3458.00  ...  3604.00  3718.0   06:54:36
 600 a  7/26/2019     NaN      NaN  ...      NaN     NaN   06:54:36
 600 b  7/26/2019     NaN      NaN  ...      NaN     NaN   06:54:36
 700 a  7/26/2019  3462.0  3684.00  ...  3821.00  3800.0   06:54:36
 700 b  7/26/2019  4290.0  4414.00  ...  4303.00  4336.0   06:54:36
 900 a  7/26/2019  2863.0  3059.00  ...  3075.00  3313.0   06:54:36
 900 b  7/26/2019  4480.0  4632.00  ...  4873.00  4843.0   06:54:36
1000 a  7/26/2019     NaN      NaN  ...  4426.00  4751.0   06:54:36
1000 b  7/26/2019     NaN      NaN  ...  4388.00  4239.0   06:54:36
6:54:40   7/26/2019   -35.0   -34.75  ...     8.75     9.0   06:54:40
 500 a  7/26/2019  3995.0  4056.00  ...  4571.00  4480.0   06:54:40
 500 b  7/26/2019  3837.0  3974.00  ...  3720.00  3619.0   06:54:40
 600 a  7/26/2019     NaN      NaN  ...      NaN     NaN   06:54:40
 600 b  7/26/2019     NaN      NaN  ...      NaN     NaN   06:54:40
 700 a  7/26/2019  3501.0  3468.00  ...  3897.00  3911.0   06:54:40
 700 b  7/26/2019  4422.0  4331.00  ...  4737.00  4505.0   06:54:40
 900 a  7/26/2019  2681.0  2749.00  ...  3375.00  3269.0   06:54:40
 900 b  7/26/2019  4542.0  4602.00  ...  4505.00  4442.0   06:54:40
1000 a  7/26/2019     NaN      NaN  ...      NaN     NaN   06:54:40
1000 b  7/26/2019     NaN      NaN  ...      NaN     NaN   06:54:40

I want to plot a values and b values in columns 2-178 on two separate plots (a plot and b plot), and I want to do this for each period of time. I'll eventually want to click through plotting each time to see the changes over time (like a plotting GUI). I need to pull out the selected columns based on time and index name for each set of timestamps. For example, I want:

a500 = [3880.0  4068.00  ...  4562.00  4398.0]
a600 =  [NaN      NaN  ...      NaN     NaN]
a700 = [3462.0  3684.00  ...  3821.00  3800.0]
a900 = [2863.0  3059.00  ...  3075.00  3313.0]
a1000 = [ NaN      NaN  ...  4426.00  4751.0]

And I want to be able to update on button click to:

a500 = [3995.0  4056.00  ...  4571.00  4480.0]
a600 =  [NaN      NaN  ...      NaN     NaN]
a700 = [ 3501.0  3468.00  ...  3897.00  3911.0]
a900 = [2681.0  2749.00  ...  3375.00  3269.0]
a1000 = [ NaN      NaN  ...      NaN     NaN]

I won't know the timestamps in advance. The structure of the rows should be consistent throughout the entire dataframe (row that starts with time and associated values, followed by alternating a and b rows, then repeat for new time value). I would like to be able to keep NaNs because these are non-zero values that I do not want to graph as zeros.

I've tried using .loc to search for rows that start with the value that I want (eg a500=data.loc['500 a'] ), but it kicks out error messages (eg KeyError: '500 a' ).

Tl;dr: need help selecting subsets of rows based on columns in a pandas dataframe as a step towards graphing.

Answer 1

It took a lot of playing around, but I did manage to get .iloc to work:

n=1
m=n+10
subdf=df.iloc[n:m]
newdf=subdf[subdf.columns[1:178].tolist()]

This solution works for me because I know this dataframe has repeating row labels and a defined number of columns. The n and m values are placeholders for when I eventually want to iteratively graph portions of my dataframe. So, as long as the number of associated rows is constant for a value (eg I have 10 rows for every new timestamp), this solution will work.