Fix drift-rate when estimating drift-criterion?

[Nicolás Sánchez-Fuenzalida]

Hi, I have a 2x2 (2 bias manipulations and 2 bias directions) experiment and I'm trying to estimate bias in drift criterion and starting point. See below the settings I'm using. As far as I know, the dc parameter should capture any bias in the drift rate. Does that mean that I don't need to set 'v' to also depend on 'bias_direction'? 

hddm.HDDMStimCoding(data, stim_col = 'stim',
                            split_param = 'v',
                            include = ('sv'),
                            drift_criterion = True,
                            bias = True,

                            depends_on = {' 'v': ['bias_manipulation'],

                                                         'z': ['bias_manipulation', 'bias_direction],

                                                          'dc': ['bias_manipulation', 'bias_direction']},

                            p_outlier = .5
                            )

[Alexander Fengler]

Hi,

some comments. 

1. Is 'stimcoding' the desired setup here? Your explanations don't let me deduce that.

2. In principle your 'depends_on' setting could be correct as is, it depends on what the manipulation does (specifically whether it is redundant with the 'stim' column). 

3. your p_outlier = 0.5 setting seems very extreme --> p_outlier = 0.05?

[Nicolás Sánchez-Fuenzalida]

Thought I sent this answer but apparently I didn't.

Thanks Alexander for picking up my question.

1. I do want to use stimcoding (as far as I know) because I need to estimate bias parameters. I want to compare bias across the two bias manipulations by comparing whether the difference between bias direction within a bias manipulation is bigger for drift criterion or starting point.

2. The manipulation is a monetary punishment scheme where you lose more money for answers A or B, and the other one is a base rate manipulation where either A or B stimuli are more prevalent. The bias directions are biased-to-A and biased-to-B. The stim column indicates whether the stim was A or B.

3. That's a typo. I meant 0.05