variable already defined error in the foreach loop

[Dominique Bourget]

Hello,
I am trying to make a use of the Python-Stata interface in Stata foreach loop. My loop is shown below:

foreach a1 of local alphaGrid {

python: a = float(Macro.getLocal("a1"))

// predict using the best value for alpha
python: mnb = MultinomialNB(alpha = a, class_prior = None, fit_prior = True)

// model Accuracy, how often is the MultinomialNB classifier correct?
python: Y_mnb_pred = mnb.fit(X_train, np.ravel(Y_train)).predict(X.iloc[:, :-1])

// right now Y_mnb_best_pred is a numpy array; so change it into a list
python: Y_mnb_pred = Y_mnb_pred.tolist()

// transfer the python variable Y_mnb_best_pred as the Stata variable 'yBestPred'
python: Data.setObsTotal(nobs)
python: Data.addVarFloat('yPred')
python: Data.store(var = 'yPred', obs = None, val = Y_mnb_pred)

generate correct yPred == category
tabulate correct
summarize correct
`accuracy' = r(mean)

replace Accuracy = `accuracy' in `i'
replace Alpha = `a1' in `i'

// update the counter i
local i = `++i' 
drop yPred 

}

and this is generating an error, the error message is "variable yPred already defined".
I thought I would be fine with this since I am doing -drop yPred- at the end of the loop?
How can I fix this error?

Thank you,

[William Lisowski]

Your code has an error that will cause it to fail the first time through the loop just after creating yPred, which then is not dropped because you don't get to the bottom of the loop. So when you rerun your code to try to fix the problem, the left-over yPred already exists.

This corrects the problem I mentioned.

local accuracy = r(mean)

[Dominique Bourget]

hello,

Thank you for your reply.

When I execute -clear- before I re-run the loop, and after changing -`accuracy' = r(mean)- to -local accuracy = r(mean)-, I am still given the same error message "variable yPred already defined"....
Is there something else that is wrong with my loop? Thank you,

[Dominique Bourget]

apparently the codes run to the error when -generate correct yPred == category- is ran for the first time. How should I fix this line? Thank you,

[Bjarte Aagnes]

generate correct yPred == category

should probably be

generate correct = ( yPred == category )

[Dominique Bourget]

Thank you!

[Bjarte Aagnes]

I suggest moving the python code to a separate py file. This also make it easy to spot the Stata syntax errors:

local i = 0

foreach a1 of local alphaGrid {

    local ++i    
    capt drop yPred

    python script mnb_pred.py  /* creates Stata variable yPred */
    
    generate correct yPred == category

    tabulate correct
    
    summarize correct
    `accuracy' = r(mean)

    replace Accuracy = `accuracy' in `i'
    
    replace Alpha = `a1' in `i'    
}

In addition, the variable name "correct" might be changed to "correct_prediction" etc. The local accuracy seems unnecessary:

su correct_prediction , meanonly
replace Accuracy = r(mean) in `i'