Following this link(http://ski.clps.brown.edu/hddm_docs/tutorial_python.html#within-subject-effects), I’m trying to run within-subject analysis, and I have a few questions regarding model specification. Here is a list of three variables that I’m interested in.
Session - 2 levels of “behav”, “fmri”
Condition - 2 levels of “Black”, “Like”
Diff (difficulty) - 2 levels of “1”, “0”
My primary goal is to compare subject performance between “fmri” and “behav” sessions. For example, for those in easy & Black condition (diff =0, condition = Black), do those in (session = behav) differ from those in (session = fmri)?
Using pseudo-code, I need an intercept of
Black_easy_behav -> black[black] : easy[easy] : behav[behav]
And second term of
Black_easy_fmri -> black[black] : easy[easy] : behav[fmri] waiting to be compared.
Using this input, I type -
m_within = hddm.HDDMRegressor(data, "a ~ C(session, Treatment('fmri')):C(condition, Treatment('Black')):C(diff, Treatment(0))")
I get this -
[‘a_Intercept’, ..., “a_C(session, Treatment(‘fmri’))[T.behav]:C(condition, Treatment(‘Black’))[T.Like]:C(diff, Treatment(0))[T.1]”]
While I have the intercept term, the second term is not what I wanted. I was looking for something like this ->
“a_C(session, Treatment(‘fmri’))[T.behav]:C(condition, Treatment(‘Black’))[T.Black]:C(diff, Treatment(0))[T.0]”]
In summary, my questions are
How do I get the specified results that I need? I’m a bit confused by how the co-variates are added. I can output the right combination (222, all8) by having
> m_within = hddm.HDDMRegressor(data, "a ~ 0 + C(session, Treatment('fmri')):C(condition, Treatment('Black')):C(diff, Treatment(0))")
Then I have this error. Also not sure how to solve.
> AssertionError: Step-out procedure failed
What would the intercept be if there are more than 1 term? In the tutorial, there intercept is only 1 variable. However, if mine has more variables for intercept, what does it become? Case 1: a ~ A:B, in this case I assume intercept is A:B. Case2: a ~ A:B + B:C, would intercept be A:B, or the combination of (A:B + B:C)
What's the difference between C(diff, Treatment(1))[T.0] vs. C(diff, Treatment(0))[0]?
Why sometimes there is [T.0], sometimes its just [0]? I’ve already made them “factors” rather than “integer/numbers” via R. Upon comparing my results and those from the Demo, I see a divergence from very early on here -
Where the Demo has
> from patsy import dmatrix dmatrix("C(stim, Treatment('WL'))", data.head(10)) and outputs all levels
> DesignMatrix with shape (10, 3) Intercept C(stim, Treatment('WL'))[T.LL] C(stim, Treatment('WL'))[T.WW]
My line of ...
> from patsy import dmatrix("C(condition, Treatment('Black'))", data.head(10)) only generates the following with out the other conditions. I hope to see something like "C(condition, Treatment('Black'))[T.Like]"> DesignMatrix with shape (10, 1) Intercept Whats more, if I type
> dmatrix("C(condition, Treatment('Like'))", data.head(10)) I got result of this. and I wasn't sure what caused the level issue.
`> PatsyError: specified level 'Like' not found
Following this link(http://ski.clps.brown.edu/hddm_docs/tutorial_python.html#within-subject-effects), I’m trying to run within-subject analysis, and I have a few questions regarding model specification. Here is a list of three variables that I’m interested in. Session - 2 levels of “behav”, “fmri” Condition - 2 levels of “Black”, “Like” Diff (difficulty) - 2 levels of “1”, “0”
My primary goal is to compare subject performance between “fmri” and “behav” sessions. For example, for those in easy & Black condition (diff =0, condition = Black), do those in (session = behav) differ from those in (session = fmri)?
Using pseudo-code, I need an intercept of Black_easy_behav -> black[black] : easy[easy] : behav[behav] And second term of Black_easy_fmri -> black[black] : easy[easy] : behav[fmri] waiting to be compared.
Using this input, I type -
m_within = hddm.HDDMRegressor(data, "a ~ C(session, Treatment('fmri')):C(condition, Treatment('Black')):C(diff, Treatment(0))")I get this -[‘a_Intercept’, ..., “a_C(session, Treatment(‘fmri’))[T.behav]:C(condition, Treatment(‘Black’))[T.Like]:C(diff, Treatment(0))[T.1]”]While I have the intercept term, the second term is not what I wanted. I was looking for something like this ->“a_C(session, Treatment(‘fmri’))[T.behav]:C(condition, Treatment(‘Black’))[T.Black]:C(diff, Treatment(0))[T.0]”]In summary, my questions are
> m_within = hddm.HDDMRegressor(data, "a ~ 0 + C(session, Treatment('fmri')):C(condition, Treatment('Black')):C(diff, Treatment(0))")Then I have this error. Also not sure how to solve.> AssertionError: Step-out procedure failedWhere the Demo has
> from patsy import dmatrix dmatrix("C(stim, Treatment('WL'))", data.head(10))and outputs all levels> DesignMatrix with shape (10, 3) Intercept C(stim, Treatment('WL'))[T.LL] C(stim, Treatment('WL'))[T.WW]My line of ...> from patsy import dmatrix("C(condition, Treatment('Black'))", data.head(10))only generates the following with out the other conditions. I hope to see something like"C(condition, Treatment('Black'))[T.Like]"> DesignMatrix with shape (10, 1) InterceptWhats more, if I type> dmatrix("C(condition, Treatment('Like'))", data.head(10))I got result of this. and I wasn't sure what caused the level issue. `> PatsyError: specified level 'Like' not foundThank you so much in advance!