Using Filters
The Datasets of your Database can be filtered by the following filters. The filters below are all optional filters. They cannot be used with the Databases view and the Datasets view.
 Date: This is the first column of each view. The name of this column might be different between views. This has a datetime type and can be used to retrieve data within a specific date range. The supported operators for the Date filter are '>' , '<' ,'>=', and '<='.
SELECT * FROM AAPL WHERE Date <= '2016224'
This column is also the only column you can use to order the data in ascending or descending order.  Collapse: Indicate the desired frequency. When you change the frequency of a Dataset, Quandl returns the last observation for the given period. For example, by collapsing a daily Dataset to monthly, you will get a sample of the original Dataset where the observation for each month is the last data point available for that month. Set collapse to one of these values: daily, weekly, monthly, quarterly, or annual.
SELECT * FROM AAII_SENTIMENT WHERE Collapse='weekly'

Transform: Perform calculations on your data prior to downloading. You can set the transform parameter to one of the following values, as shown below in SQL:
SELECT * FROM AAII_SENTIMENT WHERE Transform='rdiff'
In the following formulas, y[0] refers to the starting date of your query's result, not the underlying table. y[latest] refers to the ending date based on the other parts of your query.
Name Description Formula diff Rowonrow change. This parameter shows the difference between each column of a row and the value of that column in the previous row. y"[t] = y[t]  y[t1] rdiff Rowonrow percent change. This parameter shows the percentage difference between each column of a row and the value of that column in the previous row. y"[t] = (y[t]  y[t1]) / y[t1] rdiff_from Latest value as percent increment. This parameter shows a ratio between each column of a row and the value of that column in the latest row. y"[t] = (y[latest]  y[t]) / y[t] cumul Cumulative sum. This parameter calculates the sum of all preceding column values. y"[t] = y[0] + y[1] + ... + y[t] normalize Scale the time series such that the oldest value is 100. (y"[t] = y[t] / y[0]) * 100.