Groupby Aggregrations with NaN

[felixerdy]

I'd like to group and aggregate my dataframe. The df can have different columns like (timestamp_unix, latitude, longitude, ..., accelerometer_z, temperature)

Usually, each row has a timestamp and one value (like accelerometer, temperature, ...). All other phenomena are null. I'm grouping my df based on a geo_id I'm calculating before and would now like to calculate aggregations.

    // Specify aggregation functions for each column
    const aggregationFunctions: { [key: string]: any } = {
      timestamp_unix: 'min',
      latitude: 'max',
      longitude: 'max',
      overtaking: 'sum',
      finedust_pm1: 'mean',
      finedust_pm2_5: 'mean',
      finedust_pm4: 'mean',
      finedust_pm10: 'mean',
      distance: 'sum',
      humidity: 'mean',
      accelerometer_x: 'mean',
      accelerometer_y: 'mean',
      accelerometer_z: 'mean',
      temperature: 'mean',
    }

    // Group by geolocation and apply aggregations
    const grouped = df.groupby(['geo_id']).agg(aggregationFunctions)

    grouped.print()

Unfortunately, null cells get included in the aggregation (as 0 I think). Thus, temperature_mean does not represent the real mean temperature. Filling null cells with NaN does not work either, as the resulting aggregation values all are NaN now.

const grouped = df.fillNa(NaN).groupby(['geo_id']).agg(aggregationFunctions)

How can I exclude null or NaN values from my groupby aggregations?