Pandas Indexing

Basic

There are few method we can access our DataFrame with pandas. There are .ix, .loc and .iloc. We will examine them one by one.

.ix

Since .ix is deprecated right now. And .loc, .iloc are expected replacements. .ix is a mixture of .loc and .iloc which supports mixed integer and label based indexing. We will focus more time with .loc and iloc instead of .ix.

.loc

Label-location based indexing

It has the basic format of pd.DataFrame.loc[<row(s)>,<column(s)>]. If we want to select single row of some DataFrame we created. We can use myDataFrame.loc[0,:]. Since it starts with number 0. And colon (:) means selecting all columns. If we want to select certain rows or columns. We can use myDataFrame.loc[[0,1,2],:] or myDataFrame.loc[0:2,:] to select row 0,1,2 and all columns. And please notice that when we use 0:2 colon expression, it will include both endpoints (both 0 and 2), unlike range(0,2), only include left endpoint, not right endpoint (only 0, not 2). Or we want to select certain columns, we can use myDataFrame.loc[:,['col_name1','col_name3']] will only select two columns of ‘col_name1’ and ‘col_name2’. We also can select consecutive columns using myDataFrame.loc[:,'col_name1':'col_name3'] to select three columns of ‘col_name1’ to ‘col_name2’.

Boolean Selection

To select observations with State column equal to ‘New York’, we can use myDataFrame.loc[myDataFrame.State=='New York',:].

.iloc

Integer-location based indexing

It has the same format as .loc of pd.DataFrame.iloc[<row(s)>,<column(s)>]. To select single of multiple rows or columns, we can use myDataFrame.iloc[:,[0,3]] to select all rows and first column(0) and forth column(3). We can also use 0:3 colon expression. However in this case, it’s inclusive from left endpoint but exclusive from right endpoint like list(range(0,3)). In this case 0:3 will show only 0,1,2 columns. You can also do a boolean selection like .loc in .iloc.