Moving averages with crossfilter.js and dc.js - crossfilter

Question

I would like to use crossfilter.js with dc.js in order to represent moving averages.
Is it possible to calculate moving averages on a set of data or if the data passed in has to already have the averages calculated?
If it's possible, can you please provide a short example.
Thank you,

Answer 1

I would love to be proven wrong but I can't really think of a way to do this in crossfilter.
In the case of a timeseries, the reduce() function would be used to perform aggregations across data, grouped by date, as per the highlighted cells below:
With a moving average you'd need to perform an operation by looping down the rows:
Server-side, I wouldn't choose SQL to calculate a moving average on database data; likewise I wouldn't use crossfilter client-side.

Answer 2

I don't know if you're still after any help with this, but having just overcome this problem myself, I thought I'd share my solution in case it helps anyone else.
First I declare some pretty standard reduce functions to return a simple average, with the addition of two variables to store the averages for t-1 and t-2. I created a 3 period moving average, but if you want a longer one you could easily replace these vars with an array which you push and splice the data to/from.
function movingAveInit(){
return {
total: 0,
count: 0,
average: 0,
//These are the two vars for storing previous periods
avet1: 150000,
avet2: 150000
};
}
function movingAveAdd(p,v){
//'apples' is the property that you're interested in finding the moving average for
p.total = p.total + v.apples;
p.count = p.count +1;
p.avet2 = p.avet1;
p.avet1 = p.average;
p.average = p.total/p.count;
return p;
}
function movingAveRemove(p,v){
p.total = p.total - v.apples;
p.count = p.count - 1;
p.avet2 = p.avet1;
p.avet1 = p.average;
p.average = p.total/p.count;
return p;
}
Next call these on your group (the dimension should be whichever time period you are using):
var movingAverageGrp = yourDimension.group().reduce(movingAveAdd, movingAveRemove, movingAveInit);
Now when you build a chart with this group, add the following to its valueAccessor():
yourChart
.valueAccessor(function (d) {
var mAv = ((d.value.avet2 + d.value.avet1 + d.value.average) / 3);
return mAv;
})
If you stored previous periods in an array, you can calculate the moving average by using this function instead:
.valueAccessor(function (d) {
var sum = d.value.yourArray.reduce(function(a, b){ return a + b; });
var count = d.value.yourArray.length;
var mAv = (sum / count);
return mAv;
})
It's not a particularly beautiful solution, but it seems to work for me.