Groupby - suggestion

KNIME Analytics Platform
1

I think it would be handy to add IQR to the list of aggregation functions.

Cheers,

Steve.

2

Hi Steve,

Isn't the "Quantile" aggregation function what you are looking for? Adjusting its parameter you can obtain first, second and third quartiles.

Hope that helps,

Gio

3

That's fine for one or two columns, but suppose you have hundreds of columns - having to manually do Q3 - Q1 for each one would become tedious. One could perhaps loops through columns, but having it directly as a GroupBy aggregate function would be far more efficient and simpler.

4

In that case I think I would use Pattern Based or Type Based Aggregation in order to set Q3 - Q1 only once for a large set of columns. I think those options were thought exactly to avoid having manually setting many aggregation rules.

Cheers

5

In the GroupBy node, Pattern Based or Type Based Aggregation allows you to do Q1 and Q3, but I see no way to form (Q3 - Q1). You still have do that manually or in some sort of loop.

9

@SOH979
You can easily calculate the (Q3 -Q1) using the Math Formula (Multi Column) node, without any looping.

1 Like
10

Thanks, I hadn’t spotted this node, though I ended up just using an R Snippet node.

Cheers,
Steve.

2 Likes