Aggregation Mode
The Aggregation Mode dictates what happens when multiple raw values are encountered for a given time slice. The number of values within a time slice is determined by the Sample Size properties. For more information on how time slices are populated, see the How the Tag Historian System Works page.
Aggregation Mode  Description 

Timeweighted Average  The values are averaged together, weighted for the amount of time they cover in the interval.
Note: In scripting, Timeweighted Average is instead labeled Average. These are the same.

Min/Max  The minimum and maximum values will be returned for the window. In other words, two rows will be returned. If only one value is seen in the time slice, only one row will be returned. 
Closest Value  The value closest to the ending time of the interval will be returned.
Note: In scripting, Closest Value is instead labeled Last Value. These are the same.

Basic Average  The values are summed together and divided by the number of values.
Note: In scripting, Basic Average is instead labeled Simple Average. These are the same.

Sum  The values in the time slice are summed together. 
Maximum  The maximum value in the time slice. 
Minimum  The minimum value in the time slice. 
Duration On  Returns the number of seconds that the value was recorded as nonzero. 
Duration Off  Returns the number of seconds that the value recorded as zero. 
Count On  Returns the number of times the Tag's value went from a zero value to nonzero. 
Count Off  Returns the number of times the Tag's value changed from a nonzero value to zero. 
Count  Returns the number of times a value was recorded 
Percent Good  Timeweighted percentage of good values over the date range. 
Percent Bad  Timeweighted percentage of bad values over the date range. 
Range  Returns the range between the highest and lowest value for the period.
This feature was changed in Ignition version 8.1.17:
Range mode will return "0" if the historical tag value remains static over the given Time Range. 
Standard Deviation  Standard Deviation  Returns the standard deviation of values, or how much spread is present in the data; low standard deviation shows the values are close to the mean, and high standard deviation shows that the data points are spread out over a large range of values. Only good quality values are used when calculating 
Variance  Returns the variance of values. Similar in concept to standard deviation. Only good quality values are used when calculating. 
Sample Size
The sample size determines how many data points will be returned from the query.
On Change
An On Change query will return points as they were logged, and can be thought of as a "raw" query mode. This means that the results may not be evenly spaced. Also, it is important to note that every changed value will result in a row, and therefore if you are querying multiple tags and once, you may end up with more rows than you anticipated. For example, if Tag A and Tag B both change, you would end up with [[A_{0}, B_{0}],[A_{1}, B_{0}], [A_{1}, B_{1}]].
If you want to essentially retrieve raw values, while coalescing them down into fewer rows, try using the Interval sample mode, with an interval set to your largest acceptable time between rows, and select "prevent interpolation" from the advanced settings.
Natural
A Natural query will look up the logging rate for the queried tags (when possible), and return results spaced apart at that rate. This means that the return size will vary with the date range.
Fixed
You can use the Sample Size and Aggregation Mode on the Tag History binding type to fix the number of records that are retrieved. The Fixed sample size will cause the binding to retrieve all records from the date range, and aggregate them evenly between a fixed number of points. This will ensure that the number of rows will remain the same without regard to the size of the dataset. In windows where users are able to select a large range of data, Fixed is recommended as it will prevent the property from loading an excessive number of records.
In cases where the number of points can not evenly represent the data from the date range, an extra point will be added, making the final size of the dataset the fixed value + 1.
Selecting the Min/Max aggregation mode returns two rows of data for every row requested. Each pair represents a minimum and a maximum result from the underlying data. Therefore, a table with a fixed length, would return double the requested amount with Min/Max aggregation mode selected. With Min/Max aggregation mode selected, and with a fixed row length of one, the data set returns the oldest tag value of the time range
The following image shows a Tag History Binding pulling data from the last one day. The Sample Size is configured to Fixed with a value of 100, and the Aggregation Mode is set to Basic Average. This means that the binding will query for data from the last one day, regardless of how many records there are, and create 100 timeslices that are evenly dispersed between the start and end periods of that range. Then, a basic average of the tag values are calculated for each timeslice. The resulting values are then returned to the property.
Note the Insert Propertyicon next to Sample Size. This allows a property binding to determine the number of data points, so you could change the size to increase or reduce the amount data points on the chart from the client.
Interval
Where as the Fixed sample size will calculate time slices based on the date range, the Interval sample size allows you to determine the size of the time slices. This sample size will divide the date range by the interval size to determine the size of each slice. Because of this, it is recommended to use an interval that is evenly divisible by the date range. However, in the event that the date range is dynamic or user driven, interpolation will handle any partially built slices. Even though the binding may attempt to evenly distribute the slices, there may be an extra row that represents the current values as they are building an interval.
The image below shows a Realtime range of 60 minutes. The Aggregation Mode is set to Timeweighted Average, and the Sample Size is set to Interval for 5 minutes. This means that the binding will query for data ranging from 60 minutes ago to now (or whenever the binding last executed, in the case that polling has been turned off). That 60 minute window will be divided as evenly as possible into 5 minute timeslices, so there should around 12 timeslices. Each time slice will aggregate its value based on the timeweighted average of all values within that slice.
The example uses a Realtime range, but a Historical range could easily be used instead.
Note the Insert Propertyicon next to Sample Size. This allows a property binding to determine the number of data points, so you could change the size to increase or reduce the amount data points on the chart from the client.