How do you draw "fit" line in time series builder in kibana (7.1.1)?

[YuWatanabe]

I am trying to draw "fit" line in time series in below red points.

I have referenced the github issue but not quite sure if fitting .

Can we do the fit (average) same as timelion in timeseries builder ?

Thanks,
Yu Watanabe

[mattkime (Matthew Kime)]

Unfortunately it does not appear to be possible as stated in the github issue.

[YuWatanabe]

@mattkime
Thank you for the reply.

Reading the github conversation carefully , my understanding for fit was to make chart smoother, links to the below comment by polyfractal.

Upsampling/interpolating for visual consumption is totally acceptable imo. It's just making a chart more pleasant to look at, or to smooth out gaps so as to not distract the user. That's fine, and also totally doable client-side after the final aggregation is done running.

First I would like share my requirement. (My apology I did not share at first place.)

I have a dataset where each data (group of integer value) is sampled on random day.

Dataset will be something like,

2019-08-05 type: fruit, price: 100
2019-08-05 type: veg, price: 200
2019-08-05 type: fish, price: 300
2019-08-13 type: fruit, price: 200
2019-08-13 type: veg, price: 50
2019-08-13 type: fish, price: 300
...

For each day , I will need to calculate below metric (price in above) .

1. Max
2. Min
3. Standard Deviation (standard_deviation in extended_agg)
4. Average

To plot the data point precisely on the chart , I want to use the daily interval .

Also, to visually understand the growth between the samples, I would like to show the line chart as below instead of interpolating the data point as 0 on other days.

I assume timelion is not the option for me since the metric aggregation does not support the standard deviation .

I thought if min_doc_count option in date histogram is available in time serires builder , chart above might be available . Is min_doc_count in date histogram ?

Otherwise, I will use vega .

Thanks,
Yu

[mattkime (Matthew Kime)]

Hello @YuWatanabe

I think I'm still working to understand your question. A 'fit' line attempts to summarize a dataset as an equation. It sounds to me like you want a line chart - which you have in your screenshot.

How does your screenshot differ from your goal?

[YuWatanabe]

@mattkime

My apology for late reply and sorry for the confusion.

I have artificially connected the dots using powerpoint in the screenshot.
Usually, for sparse data , data points are not connected .

To avoid this , you will need to create date histogram aggregation which the buckets next to each other has valid values.

In my case like person who is already aware of the buckets that data does not exist, filtering out the unnecesssary buckets and just connect the lines using min_doc_count would be fantastic.

Thanks,
Yu Watanabe

[mattkime (Matthew Kime)]

I think the best route is to figure out why those 'empty' buckets are appearing. Can you share a screenshot of the config for Average/max/min/std deviation? It would also be helpful to see your field mapping.