I stumbled upon an interesting question the other day, which was; ‘how do I find the average of a range of numbers that meet criteria, and by the way, I want to exclude the outliers?’

Now, if all we needed was a simple average that met criteria we could use an AVERAGEIF or AVERAGEIFS formula.

Or if we just wanted to exclude the outliers we could use the TRIMMEAN function which returns the mean (average) of the interior proportion of values.

The syntax is:

TRIMMEAN(array, percent)

Where the ** array** is the range containing your values and

**is the fractional number of data points you want to exclude from the top and bottom of your data set.**

*percent*But we want to effectively find the AVERAGEIF ** and** TRIMMEAN.

Unfortunately this pair don’t get on so well in the same cell so I’ve got an alternative and it’s not called TRIMMEANIF….but maybe there’s an idea.

*Note to self; write to Microsoft suggesting new TRIMMEANIF function!*

Ok, let’s take this list as our example:

And let’s say we want to find the average Units for Size 8 and exclude the top and bottom 40%.

*Note: you wouldn’t normally exclude 40% of your sample but since mine is so small (so it’ll easily fit on the screen) I have had to make the proportion to exclude quite large.*

Here’s the formula:

=TRIMMEAN(IF(B2:B12=8,C2:C12),40%)

It’s an array formula so you need to enter it with CTRL+SHIFT+ENTER.

In English it reads: Find the MEAN (average) of the values in the range C2:C12, where the size in the range B2:B12 is 8, oh, and by the way, ignore the top and bottom 40% of the values.

My nested IF formula returns an array of Units that are size 8, and if we step through how the formula evaluates we can see it looks like this:

Step 1: Evaluate the IF formula logical test and expand the range C2:C12

=TRIMMEAN(IF({TRUE;TRUE;TRUE;TRUE;TRUE;FALSE;FALSE;FALSE;FALSE;FALSE;FALSE}, {2;4;5;6;9;4;10;8;9;4;1}), 40%)

Step 2: create an array of the values that evaluate to TRUE

`=TRIMMEAN({2;4;5;6;9}, 40%)`

Step 3: strip out the outliers from the array of values

`=TRIMMEAN({4;5;6})`

Step 4: find the mean (average) of the remaining values

=5

Note: if you run this formula through the Evaluate Formula tool you will see it work through the steps above.

## Average Top 3 Values

Another member of the Average family that you might find useful is the AVERAGE LARGE formula which finds the average of the top 3 values.

## Spread the Word

If you liked this or know someone who could use it please click the buttons below to share it with your friends on LinkedIn, Google+, Facebook and Twitter.

And if you want to help start a movement to get Microsoft to include a new TRIMMEANIF function in the next release of Excel click on all of them! Heck, why not click on all of them anyway and have a clickety click day 🙂

Debbie Dyche says

How do I get an average cost per sale from a pivot table??? I can’t find anything on averages for pivot tables.

Mynda Treacy says

Hi Debbie,

I’m assuming you’re talking about regular PivotTables and not Power Pivot? If so then you could try using a calculated field:

http://www.myonlinetraininghub.com/excel-pivottable-calculated-fields

Note: you should know that a PivotTable aggregates the data and then performs the calculation on the values in each row. I assume you will have columns for Total Costs and Count of Sales. Depending on the level of detail you drill down to in your row labels, dividing Total Costs by Count of Sales may, or may not give you the average you had in mind.

Mynda

Duncan says

I like this blog! It reminds me of what happens in my training room. I often say, you ask me questions that I never think about because I can’t walk in your shoes. Then I enjoy solving their problems.

The same here, lots of interesting problems and solutions to keep the grey cells ticking. Not always earth shattering but then they don’t have to be, do they?

Duncan

Mynda Treacy says

Cheers, Duncan. Glad to help keep your grey cells alive 🙂

Leslie Glasser says

Although Excel offers the TRIMMEAN function, it does not seem to be the statistically best measure of ‘central tendency’ since it relies on arbitrary selection of the proportions of the data to be excluded and (worse) assumes that the data spread is symmetrical.

A more satisfactory measure is MEDIAN which is the central value (or mean of two if there is a even number of data points), thus automatically excluding outliers on both ends of the data. Technically, it is a more ‘robust’ measure than TRIMMEAN.

Mynda Treacy says

Thanks for your wisdom, Leslie 🙂

If you’d like to calculate the MEDIAN IF the formula would be:

Entered as an array formula with CTRL+SHIFT+ENTER.

Kind regards,

Mynda.