modelling factor interactions

[Tobias Hauser]

Dear all

I have a design with 2 factors (A, B) and there is an interaction of them on RT and accuracy. I.e. under condition A==1, the effect of B is an increase in RT, whereas if A==2, then B causes a decrease in RT. Is there a good way to model that for a drift rate/etc?
I could just parameterize the whole into one factor using the interaction (i.e., B(x|A=1)=[1 2 3], B(x|A=2)=[3 2 1]), but this omits any offset differences between the conditions. Is there any other way to model it?
Ideally, my model would test the effect of condition A, B, and their interaction A*B on drift rate/etc.
Thanks a lot for your help!

Best,
Tobias

[Sam Mathias]

Assuming your design is completely within-subjects, use the HDDMRegressor class:

http://ski.clps.brown.edu/hddm_docs/tutorial_python.html#within-subject-effects

Since that module takes patsy formulae, you can easily code interaction terms as well as main effects. Assuming there are only two levels of A and B, something like 'v ~ 0 + C(A) * C(B)' should do what you want.

Samuel Robert Mathias

Postdoctoral Associate
Department of Psychiatry, Yale School of Medicine
Suite 3014, 2 Church Street South
New Haven, CT 06519
http://www.srmathias.com

--
You received this message because you are subscribed to the Google Groups "hddm-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to hddm-users+...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Tobias Hauser]

Thanks Samuel! I was also thinking of using HDDM-Regression, but wasn't sure whether there is a more straight forward way to doing it...
Best,
Tobias

[Chuan-Peng Hu]

Dear all, 

I encountered a similar problem recently, I am new to python, sorry if my question is too sample.

I also have a 2 by 2 within-subject design. let's say that one is called "moral", the other is "id" .

As suggested by Samuel, I used follow code:

m_within_subj_L = hddm.HDDMRegressor(datL_match2, "v ~ C(id)*C(moral)", include = ['t','a', 'v','z'],keep_regressor_trace=True, p_outlier = 0.05)

After running the above code, following was shown:

Adding these covariates:

 ['v_Intercept', 'v_C(id)[T.self]', 'v_C(moral)[T.moral]', 'v_C(id)[T.self]:C(moral)[T.moral]']

then I run the following code

# estimate the posterior, sampling 5000 times, discard the first 200

m_within_subj_L.sample(5000, burn=200)

Then I tried to get the parameter "v" for each:

v_Intercept,v_id,v_moral,v_id_moral = m_within_subj_L.nodes_db.ix[['v_Intercept', 

                                                                 'v_C(id)[T.self]', 

                                                                 'v_C(moral)[T.moral]', 

                                                                 'v_C(id)[T.self]:C(moral)[T.moral]'],'nodes']

but get an error:

v_Intercept,v_id,v_moral,v_id_moral = m_within_subj_L.nodes_db.ix[['v_Intercept', 

                                                                 'v_C(id)[T.self]', 

                                                                 'v_C(moral)[T.moral]', 

                                                                 'v_C(id)[T.self]:C(moral)[T.moral]'],'nodes']

Traceback (most recent call last):

  File "<ipython-input-63-9c984221c78f>", line 4, in <module>

    'v_C(id)[T.self]:C(moral)[T.moral]'],'nodes']

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\core\indexing.py", line 75, in __getitem__

    return self._getitem_tuple(key)

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\core\indexing.py", line 784, in _getitem_tuple

    return self._getitem_lowerdim(tup)

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\core\indexing.py", line 908, in _getitem_lowerdim

    section = self._getitem_axis(key, axis=i)

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\core\indexing.py", line 1018, in _getitem_axis

    return self._get_label(key, axis=axis)

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\core\indexing.py", line 93, in _get_label

    return self.obj._xs(label, axis=axis)

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\core\generic.py", line 1740, in xs

    return self[key]

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\core\frame.py", line 1997, in __getitem__

    return self._getitem_column(key)

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\core\frame.py", line 2004, in _getitem_column

    return self._get_item_cache(key)

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\core\generic.py", line 1350, in _get_item_cache

    values = self._data.get(item)

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\core\internals.py", line 3290, in get

    loc = self.items.get_loc(item)

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\indexes\base.py", line 1947, in get_loc

    return self._engine.get_loc(self._maybe_cast_indexer(key))

  File "pandas\index.pyx", line 137, in pandas.index.IndexEngine.get_loc (pandas\index.c:4154)

  File "pandas\index.pyx", line 159, in pandas.index.IndexEngine.get_loc (pandas\index.c:4018)

  File "pandas\hashtable.pyx", line 675, in pandas.hashtable.PyObjectHashTable.get_item (pandas\hashtable.c:12368)

  File "pandas\hashtable.pyx", line 683, in pandas.hashtable.PyObjectHashTable.get_item (pandas\hashtable.c:12322)

KeyError: 'nodes'

I also tried this code as suggested in this post 

v_Intercept,v_id,v_moral,v_id_moral = m_within_subj_L.nodes_db.node[['v_Intercept', 

                                                                 'v_C(id)[T.self]', 

                                                                 'v_C(moral)[T.moral]', 

                                                                 'v_C(id)[T.self]:C(moral)[T.moral]'],'nodes']

but also returned an error:

v_Intercept,v_id,v_moral,v_id_moral = m_within_subj_L.nodes_db.node[['v_Intercept', 

                                                                 'v_C(id)[T.self]', 

                                                                 'v_C(moral)[T.moral]', 

                                                                 'v_C(id)[T.self]:C(moral)[T.moral]'],'nodes']

Traceback (most recent call last):

  File "<ipython-input-65-06fbb44ee34d>", line 4, in <module>

    'v_C(id)[T.self]:C(moral)[T.moral]'],'nodes']

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\core\series.py", line 624, in __getitem__

    return self._get_with(key)

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\core\series.py", line 637, in _get_with

    return self._get_values_tuple(key)

  File "C:\Program Files\Anaconda2\lib\site-packages\pandas\core\series.py", line 682, in _get_values_tuple

    raise ValueError('Can only tuple-index with a MultiIndex')

ValueError: Can only tuple-index with a MultiIndex

Not, I am not sure how to get the value of parameter.

[Thomas Wiecki]

Can you try replacing .ix with .loc? If that doesn't help, what does m_within_subj_L.nodes_db.columns say?

--

You received this message because you are subscribed to the Google Groups "hddm-users" group.

To unsubscribe from this group and stop receiving emails from it, send an email to hddm-users+unsubscribe@googlegroups.com.

[Chuan-Peng Hu]

Hi, Thomas,

Thank you for your quick response.

I tried .loc, it returns "KeyError: 'the label [nodes] is not in the [columns]'"

the m_within_subj_L.nodes_db.columns , it says: "IndexError: only integers, slices (`:`), ellipsis (`...`), numpy.newaxis (`None`) and integer or boolean arrays are valid indices"

To unsubscribe from this group and stop receiving emails from it, send an email to hddm-users+...@googlegroups.com.

[Micah Johnson]

Hi,

Perhaps this may be of help to some...I've been playing around with different combinations, and here's what I've found (using Python 2.7, IPython and Anaconda 4.2.0)

## note: (nothing worked if I first entered, %matplotlib inline , in command line, no idea why)

v_test = test.nodes_db.ix[["v_C(T, Treatment('b'))[T.t]:C(S, Treatment('LO'))[T.LP]"],'node']

    # works! allows hddm.analyze.plot_posterior_nodes(v_test)

v_test2 = test.nodes_db.loc[["v_C(T, Treatment('b'))[T.t]:C(S, Treatment('LO'))[T.LP]"],'node']

    # works as well

v_test3 = test.nodes_db.node[["v_C(T, Treatment('b'))[T.t]:C(S, Treatment('LO'))[T.LP]"]]

    # works as well

v_test4 = test.nodes_db.node[["v_C(T, Treatment('b'))[T.t]:C(S, Treatment('LO'))[T.LP]"],'node']

    # # ValueError: Can only tuple-index with a MultiIndex

## note: above code also works for me when doing multiple at a time, e.g. v_test1, v_test2 = test.nodes_db.ix[["patsy string1", "patsy string2"],'node']

## also, not 100% sure, but for me it seemed like none of this code worked if I was trying to apply it to a concatenated model of multiple chains

# in case anyone needs to know how to save output...

hddm.analyze.plot_posterior_nodes(v_test)

plt.xlabel('v')

plt.ylabel('post prob')

plt.title('woohoo!')

plt.savefig('v_test.jpg')

Cheers!

Micah

[Micah Johnson]

Also, Chuan-Peng, I just noticed that you have a typo in your code... should be 'node' not 'nodes', I believe

[Chuan-Peng Hu]

Hi, Micah,

Many thanks for pointing out the typo! I am so embarrassed.

After correcting the typo,  m_within_subj_L.nodes_db.loc and m_within_subj_L.nodes_db.ix both works!!

Thank you again, and apology for my typo.