THW - pandas

• In ipython you can use _ to process the last thing you processed, not a great thing to use.
• Can pass pandas objects to numpy

General

• Last value for slices are not excluded, unlike numpy
• mfs.loc[x,y]: returns x through y, including y
• mfs.iloc[x,y]: integer location, does exclude y
• Uses filetype hdf5 to hold data (hdf-store), can store multiple dataframe within the same file, retrieved with keys

Dataframes

• Note: you can read excel files via a dataframe
• Can call a column by name df['column_name']
• You can’t call df['row_num'] to get a row, you have to use df.loc[row_num]
• df.ix allows you to use mixed labels or integer locations, if you have a label that is an integer, it will always default to the label that is the integer, not the integer location.
• Generally don’t want to use df[column][row] there are many other options (ix)
• Can change the indexing of a dataframe df.index = [] this will change the index to whatever labels that you want
• Tools for fixing datasets, df.fillna[x] will fill any NaN’s with the value x. Need to include df.fillna[x, inplace=True] to actually overwrite the objects.
• Most functions return a new object, the inplace=True flag overwrites the old object.
• groupby('column') make dataframes out of data that have the same value in ‘column’. Can group by two different things. If you do that, you’d have to call the row with a touple x.loc[(x,y),z]
• foo.xs returns a cross section of your data for a given index, at a level. Syntax fullhistogram.xs(x, level=1) level indicates the level of index, if you only have one there is one, if you have multiple indexs, you can pick.