Latest n diagonal in heatmap()

[kennethshsu]

Is there an easy way to heatmap() only the latest n diagonal? I supposed we can add in another parameter, I propose last_n_diagonal, default to None?

[jbogaardt]

Yes, you can use the Development estimator to omit any link ratios and those should propagate to the heatmap.

import chainladder as cl
cl.Development(n_periods=3).fit_transform(cl.load_sample('raa')).link_ratio.heatmap()

[kennethshsu]

Ahh, similar solution to what I came up with.

raa = cl.load_sample("raa")
cl.Development().fit_transform(raa[raa.valuation.year >= raa.valuation_date.year-3]).link_ratio.heatmap()

I think I am getting better at this. Only problem is now we can't see the older link ratios.😅

[jbogaardt]

Cool. By your approach you don't even need the Development estimator:

raa = cl.load_sample('raa')
raa.link_ratio[raa.link_ratio.valuation.year >= raa.valuation_date.year-3].heatmap()

[kennethshsu]

Yes, even better!

[jbogaardt]

Should this be reclassifed discussion forum Q&A?

[kennethshsu]

Yes, unless you think it's useful to add an argument to only heatmap the last n diagonals? It's a bit better as we can see the older diagonals that are not heat-mapped. But it will mean that it's another piece of code that we have to maintain. What do you think?

[jbogaardt]

I fear n_diagonals will be short lived. Eventually someone will want the same but with drop_hi and drop_low. I dont want to create more args for those use cases.

Maybe it's better to create a Boolean arg in heatmap that allows the dropped link ratios from the Development estimator to be visible or not.

[kennethshsu]

That's actually a great idea. I like that.

With this we should be able to see if the remaining LDFs have any patterns.