Statistics does not seems to work if attribute is used in an entity

[r-jean-pierre]

Hi!

Quite recently, I got new sensors and started to toy.
I just discover something that I don't understand about the "statistic" feature
Here are tiny examples:

cards:
  - type: custom:plotly-graph
    entities:
      - entity: 'climate.heaterentrance'
        attribute: 'current_temperature'
    hours_to_show: 7d

works obviously perfectly (just to show there is data):

Same with stats:

- type: custom:plotly-graph
    entities:
      - entity: 'climate.heaterentrance'
        attribute: 'current_temperature'
        statistic: mean
        period: hour
    hours_to_show: 7d

Fails: empty:

If I take another sensor whose values are not behind an attribute, then it works

  - type: custom:plotly-graph
    entities:
      - entity: 'sensor.ths_entrance_living'
    hours_to_show: 7d

Obviously:

And with stats

- type: custom:plotly-graph
    entities:
      - entity: 'sensor.ths_entrance_living'
        statistic: mean
        period: hour
    hours_to_show: 7d

This one works

Can someone explain what is my error?

[dbuezas]

Home assistant decides what statistics to use based on the sensor_class. I assume this may be the issue here.
You can check the sensor_class in http://homeassistant.local:8123/developer-tools/state

I know that sensor_class: measurementwill store mean, min and max. others store the sum and other store nothing.
As per the attributes, they don't land in the statistics db afaik.

[r-jean-pierre]

Hey, thanks for the answer.
So, the main device has no "class",
2024-02-14 09_57_23-Window
I started to investigate and try to find some workaround

First attempt was to create a dedicated sensor which only extract this attribute and make it "like the other working temperature sensors". I used the integration platform: attributes which "automatize" this work and I got sensor.heaterentrance_current_temperature

It seems to be consistent with your remark and also seems to work for HA:

But still, nothing in my custom:plotly-graph

Other attempt was to use custom:apexcharts-card (but I don't want... all my graphs are in plotly...), and both solutions works:

series:
      - entity: 'climate.heaterentrance'
        attribute: 'current_temperature'
        group_by:
          duration: 1h
          func: avg
      - entity: 'sensor.heaterentrance_current_temperature'
        group_by:
          duration: 1h
          func: avg

Any clue about it? I also feel is HA related, but dammed.... HA is able to send a rocket on the moon, but since I started to geek with it some years ago, I think it's not the first time where I cannot do straightforward out-of-the-box select avg(value) from my sensor groupby something or even more easy things...

[dbuezas]

AFAIK, there is no retroactive statistics. As to waiting time, to be honest I don't know but it wouldn't surprise me.
Although this pic of yours seems to indicate that there are some stats stored

I suggest you:

• try different periods
• double check against the inbuilt statistics card

[r-jean-pierre]

So, finally it came! But with a huge delay + shift

The blue curve is the first "fake" sensor I started to play with.
For the others ones, I need to assign one by one them the state_class: measurement and wait +/- 2 days to let HA understand to make stats over them....

So sorry for the 2 topics in the same issue. Lesson learnt:

• don't insist with statistic+period using attribute
• extract attribute into a new sensor and make sure state_class: measurement
• wait a looooooooooooong time and fully reboot HA (<-maybe this one is placebo, but I did it a lot :) )

Thank you for the guidance!

[rrozema]

Recorder is the component that aggregates the entities' state values into the statistics. It does this every 5 minutes for the short term statistics (in the statistics_short_term table) and every hour (in the statistics table) for the long term statistics. The short term statistics -those with a granularity of 5 minutes- get deleted after a couple of days, the long term statistics -with a granularity of an hour- are kept forever.
Which entities get aggregated, and how they get aggregated is determined by the value of the "state_class" attribute on each entity. You can view the entities plus their attributes in 'Developer tools' -> 'STATES'.

Entities with state_class: measurement

will be aggregated into a statistic with has_mean = 1, and will get their [min], [max] and [average] columns filled.

Entities with state_class: total or state_class: total_increasing

will be aggregated into a statistic with has_sum=1, and will get their [state] and [sum] columns filled.

All other entities

will not be aggregated into a statistic. Attributes can't have attributes themselves so these are not aggregated into statistics either.

---

To have a statistic aggregated for a numeric value stored in an attribute you can create a template sensor that has the desired state_class attribute. Please find two examples below, one for an increasing entity -like for example energy used-, plus another for an entity that represents a measurement (min, max, avg) -like for example instantaneous power, or a temperature-. The entities created for such template sensors will get aggregated into the statistics.

template:
  - sensor:
      - name: "my entity's attribute 1"
        unique_id: "124cb077-a937-4fa5-98c4-d80d57677a77"
        unit_of_measurement: kWh
        device_class: energy
        state_class: total_increasing
        state: >
          {{ state_attr('sensor.my_entity', 'attribute_1') }}
        availability: >
          {{ has_value('sensor.my_entity.attribute_1') }}

      - name: "my entity's attribute 2"
        unique_id: "124cb077-a937-4fa5-98c4-d80d57677a78"
        unit_of_measurement: kW
        device_class: power
        state_class: measurement
        state: >
          {{ state_attr('sensor.my_entity', 'attribute_2') }}
        availability: >
          {{ has_value('sensor.my_entity.attribute_2') }}

[dbuezas]

Thanks for the detailed write down @rrozema!
I hope you don't mind, I took the freedom of highlighting the core information with titles.
There is also this section in the docs for anybody else who may come across this.

[r-jean-pierre]

Thank you for the deep and explanation. Today, it sounds obvious to me super cause I found alone how to manage it, and soon or later you write exactly what you had writte. The bad thing about HA is probably everything seems to be "smooth" for the core team, but clearly not for final (non-ninja) users, and sometimes developers themselves (see for ex https://github.com/alexarch21/history-explorer-card)
I also have a post about state_class: total_increasing which was the devil for me for several days until I finally created my own fake-sensor to display it properly [/discussions/81].
My template file of fake-sensors is really dirty... Hopefully I don't have to manage an increasing number of devices everyday, otherwise, I might not have used HA at all

Hope it will help others!