Added example for concat method

[percygautam]

Description

Fixes #552

Checklist

• Does the PR follow official
  PR format?
• Has included a sample plot to visually illustrate the changes? (only for plot-related functions)
• Is the new feature properly documented with an example?
• Does the PR include new or updated tests to cover the new feature (using pytest fixture pattern)?
• Is the code style correct (follows pylint and black guidelines)?
• Is the change listed in changelog?

[OriolAbril]

Some comments.

InferenceData should always be formatted the same way.

Also, I think that using from_dict instead of convert_to_inference_data will be more concise (no meed for chain and draw coordinates and xarray import won't be needed anymore).

[OriolAbril]

Can you still use multidimensional data and a custom coord for the third dimension?

[percygautam]

@OriolAbril Do I add more examples regarding these? I don't understand the question.

[OriolAbril]

In the first versions of the example, the inference data created had groups with several variables and each of these variables had multiple dimensions. Now it looks like the generated inference data has groups that contain only a scalar?

[percygautam]

Okay, I understand now. So, you're asking for something like this, right? (I don't think coords is necessary)

data = {
         "a": (["chain", "draw", "a_dim"], np.random.normal(size=(4, 100, 3))),
         "b": (["chain", "draw"], np.random.normal(size=(4, 100))),
     }  
coords={
        "chain": (["chain"], np.arange(4)),
        "draw": (["draw"], np.arange(100)),
        "a_dim": (["a_dim"], ["x", "y", "z"])
    }
dataA = az.from_dict(posterior=data, coords=coords)
dataB = az.from_dict(prior=data, coords=coords)
az.concat(dataA,dataB)

[OriolAbril]

Yes, something like this

[percygautam]

I have used the same example as above. Should we also add examples for cases for concatenating over chain or draw?

[OriolAbril]

I tried this code locally and it does not create an InferenceData object properly. ArviZ idiom should be followed:

data = { 
   "a": np.random.normal(size=(4, 100, 3)),
   "b": np.random.normal(size=(4, 100))
}
coords = {"a_dim": ["x", "y", "z"]}
dataA = az.from_dict(data, coords=coords, dims={"a": ["a_dim"]})

[percygautam]

Thanks, for the correction. It seems the xarray dataset (used in the previous case) map dims automatically, but we need to specify the mapping in case of from_dict. Is this the case?

[OriolAbril]

xarray's Dataset allows several notations, in the one used, the data and coordinates are specified as dicts of tuples, where the tuple contains both the data and the dimensions. ArviZ uses only one syntax, dicts of arrays for all 3 of them, the values of the dicts are directly understood as data, coordinate values or dimension list

[percygautam]

Okay, thanks for the clarification :)

[OriolAbril]

This can be in the same line, it is not too long

[OriolAbril]

We should follow black formatting, the } should be on the same level as the d in data and the two lines in between indented 4 spaces.

[OriolAbril]

Thanks! Do you plan to add the concatenation over chain or draw here or in another PR? I don't want to merge before you have finished.

[percygautam]

Hey @OriolAbril , Sorry for the delay. I was on the move. I'll add examples for chain and draw.

[OriolAbril]

Don't worry, there is no rush!