Tolu Supervision Meeting - 17th September
===
###### tags: `Topic A` `Topic B` `Topic C`

:::info
- **Location:** Zoom
- **Date:** Friday 17th September, 2021 11:00 AM
- **Attendees:** Tolu, Joe, Matt
- **Apologies:** None
:::

:::warning
Why do we have this template?

- Provide a structure for our meetings.
- Provide a way for us to collaboratively note-take during the meetings.
- Give a nice summary of the discussion for anyone unable to attend the meeting.
- Give some content which can then be attached to the ePortfolio entry for the meeting.
:::
---

# Updates since the last meeting.
:::info
Jot down 3-4 bullet points as updates on progress since the last meeting.

This can replace "for info" items which you would like to inform us of, but we do not need to directly discuss.
:::

- [ ] A
- [ ] B
- [ ] C
- [ ] D

# *"The Worry List"*

:::info
- Jot down a list of any outstanding issues or queries.
- If you are able to, give an estimate of how much of our meeting time you would like to dedicate to this.
- Order your items in priority order. This will allow us to make sure we cover off the most important issues first.
:::

- [ ] A (5 minutes)
- [ ] B (10 minutes)
- [ ] C
- [ ] Machine.

# Short-term plans for next steps
:::info
- Jot down what you plan to work on next.
- The timescales will typically be between now and the next meeting.
- This can be high level. We'll prioritise these items on the call, change some of them, and we'll put them in the Actions list.
:::

- [ ] Generate data for numerical, percentage change and moving average and description for the graphs.
- [ ] B
- [ ] C
- [ ] Thinking about visualisation of missing points (using `imputeTS` as inspiration)

# Deadlines on the horizon

:::info
We will update this section periodically, so we are all aware of any upcoming deadlines. This can include publication submissions, University deadlines, or other important dates.

The longer ahead we can have visibility of these deadlines, the better we're able to support you (e.g. by reviewing drafts).
:::

- Oct 30, 2021 https://sites.google.com/view/iotda
- Deadline TBC https://uksystems.org/ UK Systems workshop
- Abstract only, a couple of hundred words.
- [31st March](https://www.youtube.com/watch?v=XhulR_kJf7Y) 2021: Submit interim report.
- 14th Month Year: Submit camera-ready version of workshop paper.
- 17th Month Year: Deadline for conference submission.

# Meeting notes

:::info
**Please note:** We will populate this section collaboratively during the meeting.
:::

- NAs removed has index 0..150 because it's a 1:1000 reduction and the points have been removed.
- Visualisations
- 'nocb', 'mean', 'mode', 'median' don't seem to be competitive
- What about compute?
- "Z score" in first figure should just be the normal value
- Different scales for the different measures
- Comparing the output between the different interpolations and different similarity measures should be on the same scale using a Z-score.
- 'locf' vs. 'nocb'
- You know the procedure by which they are missing so this is how we're going to get it back.
- We talked about tolerance windows where we remove any point that didnt go outside of this tolerance window. if we know that's why things were missing, they were suitably close to the last value we saw.
- `mae` vs `rmse`
- RMSE much more used as standard.
- Joe's external does not like MAE.
- Surely there's no difference?!
- What RMSE is, is where you have your error but it's a weighted average where the weighting is the size of the error.
- Stops you from being overinfluenced by data points on a small scale where the differences are tiny.
- RMSE weights errors (error times error) gives a weighting.
- For MSE/RMSE, the scale of the rror matters more. They are weighted by the magnitude of the error (relative to either the smaller or larger values).
- MAE is directly interpretable ("on average i am this far out")
- RMSE is less easy to interpret in the real world but it's a better metric.

# Actions
:::info
All of our actions should have the following things:
- An owner, the person leading on the activity.
- Sufficient detail so we can remember what we are supposed to do.
- A deadline for completion.
:::
:::info
These actions will appear in the next meeting notes either as an update since last time (if completed), on the worry list (if we have run into issues), or in the short-term plans (if we've had to push this item forward).
:::

- [ ] Chase NUIT about machines (Matt, Deadline)
- [ ] Ask Steve McGough about RStudio workstation access (Matt, Deadline)