New function for visual check in families between detections and parents.

ikahbasi commented 5 years ago

What does this PR do?

Function to plot a graph to show all detections in one day, and comparison between them. Plot detections for each family with separate colors.

Why was it initiated? Any relevant Issues?

PR Checklist

[x] develop base branch selected?
[x] This PR is not directly related to an existing issue (which has no PR yet).
[x] All tests still pass.
[x] Any new features or fixed regressions are be covered via new tests.
[x] Any new or changed features have are fully documented.
[x] Significant changes have been added to CHANGES.md.
[ ] First time contributors have added your name to CONTRIBUTORS.md.

calum-chamberlain commented 5 years ago

Can you put the matplotlib imports back as they were please. They need to be like like for stable imports and backend management.

ikahbasi commented 5 years ago

Can you put the matplotlib imports back as they were please. They need to be like like for stable imports and backend management.

Hi, Yes, i put them back.

ikahbasi commented 5 years ago

I'm not really sure what the use-case is for this, and why it should be included as an included plotting function in EQcorrscan. Can you justify why we should include this and continue maintaining it please?

In my opinion, visual output is useful for understanding. In this plot we can find how many events are detected by each template, which one detects more and which one detects less, density of detection in time and etc in light, fast and simple plot. Maybe we can find our mistakes and solve them. It's interesting to compare detection in cccs plot and text output of match_filter. It will be good if all examples in EQcorrscan tutorial have visual output, like obspy tutorial. And EQcorrscan will include more function for visual check of output.

calum-chamberlain commented 5 years ago

So is the purpose of this different to the cumulative detection plotting already included? I agree that plots are helpful, I just don't know where you would call this from, and what different/better information it gives than the cumulative plot, which is supported by Party.plot()?

ikahbasi commented 5 years ago

So is the purpose of this different to the cumulative detection plotting already included? I agree that plots are helpful, I just don't know where you would call this from, and what different/better information it gives than the cumulative plot, which is supported by Party.plot()?

You right. I'm not sure. Maybe this function is same as cumulative detection plot or even less useful. Maybe it is another aspect of this mean. If I'm being honest, in my case i had a simple mistake in my code then i detect duplicate of one event. false

after i saw this graph i found i had 2 events but only one point in figure. advance

I also noticed an issue in v0.3.3 that I don't know if there is in the new version or not. I make a template of a specific day, then i use the template for this day to detect event. but i did not see any detection or even self detection in cccs plot in match_filter function! cccsum_plot_2014-10-10T10_49_11_2014-10-10

but after i draw my plot, i found there is self detection in detection and even more! advance

calum-chamberlain commented 5 years ago

That is strange that the detection spikes were not showing up in the matched-filter plot - are you sure the detection plot is for one of the templates that was on that day?

I think the scope of this plot is very similar to the cumulative detection plots, but, those plots do not show the template time, which would be very helpful. I suggest that you work on adding template times into that plot, with your vertical lines as here. Personally I find the little crosses close to one another difficult to see, whereas an increasing cumulative count with time I find easier to read.

ikahbasi commented 5 years ago

That is strange that the detection spikes were not showing up in the matched-filter plot - are you sure the detection plot is for one of the templates that was on that day?

I saw this in some case, i will check more for better explanation with some example and details and i'll tell you later.

I think the scope of this plot is very similar to the cumulative detection plots, but, those plots do not show the template time, which would be very helpful. I suggest that you work on adding template times into that plot, with your vertical lines as here. Personally I find the little crosses close to one another difficult to see, whereas an increasing cumulative count with time I find easier to read.

Yes, although i use this plot in interacting mode and zoom on for better view, then i didn't notice crosses close together. You are right, if i save output of this graph in an image, it's useless. specially in case of numerous ditections, except legend of plot. Ok, it's good suggestion. i will work on cumulative detection plot and add some feature to it. Although i like origintime_detections in some cases for myself :) would you like to close this PR?

calum-chamberlain commented 5 years ago

Yup, I would go for closing this and opening a new PR for the cumulative plot.

eqcorrscan / EQcorrscan