Frequency conversion provides basic conversion of data using the new frequency intervals and allows the filling of missing data using either NaN, forward filling, or backward filling. More elaborate control is provided through the process of resampling.
Resampling can be either downsampling, where data is converted to wider frequency ranges (such as downsampling from day-to-day to month-to-month) or upsampling, where data is converted to narrower time ranges. Data for the associated labels are then calculated by a function provided to pandas instead of simple filling.
To demonstrate upsampling, we will calculate the daily cumulative returns for the MSFT stock over 2012 and 2013 and resample it to monthly frequency. We will examine the return calculation in more detail in Chapter 5, Time-series Stock Data, but for now, we will use it as a demonstration of the mechanics of up and down resampling of time-series data.