Daily mean of a sensor

Configuration
1

Hi all.

I have a sensor returning a sequence of measurements (in more or less regular intervals) such as:

0 0 0 1 1 2 3 3 3 3 4 3 2 2 2 1 0 0 0 0 0 0 1 1 2 2 2 3 ...

and I want to have a “dynamic” mean value in the non-zero data points (i.e. zeros are ignored).
The mean should bee like (sum/count):
- - - 1 2/2 4/3 7/4 10/5 13/6 16/7 20/8 23/9 25/10 27/11 29/12 30/13 30/13 30/13 ... 30/13 31/14 32/15 34/16 36/17 38/18 41/19 ...

and this computed on a daily basis, meaning that the statistics will be cleared at 0:00.

I was able to make a template sensor filtering out the zero values (hopefully, it is not wrong):

   - sensor:
     - name: data_noZero
       state: >
              {% if states('sensor.data') | float(0) > 0 %}
                 {{ states('sensor.data') }}
              {% else %}
                 {{ this.state }}
              {% endif %}

But I don’t know how to make the rest of the computation.
Using:

   - platform: statistics
     name: "data_stats"
     entity_id: sensor.data_noZero
     state_characteristic: mean
     max_age:
       hours: 24

gives me value different than when computed manually.

How to solve this?

Thank you very much for any help,
DT

2

Hi all,
after some experimenting, it seems that this modification did the trick:

   - sensor:
     - name: data_noZero
       state: >
              {% if states('sensor.data') | float(0) > 0 %}
              {{    states('sensor.data') }}
              {% else %}
              {{    this.state }}
              {% endif %}
       attributes:
         last_changed: >
              {% if states('sensor.data') | float(0) > 0 %}
              {{    now() }}
              {% else %}
              {{    this.attributes.last_changed }}
              {% endif %}

I.e. it modifies state every time a non-zero value arrives. In this way, the sensor seems to propagate updates (due to attribute.last_changed change) even when the same values following each other are measured by sensor.data.

Still, it may be wrong. So I will appreciate a more correct solution anyway …

Thanks,
DT