THIS IS A TEST FILE ... it are full of Villa thangs ....
Data are from the 'f8' and 'h1/h2/b' sets, '0000_m' files. These are
summarized in
Second Session, First (Prompted) Condition:
G1 G2 G3
Second Session, Second (Instructed) Condition:
G1 G2 G3
Strategy and accuracy data from these files are themselves summarized in
components_model_sim_sum.
---
Accuracy Chi-Squares
Empirical Data
Accuracy data are summarized here. Correct recall
versus total times a strategy was used are collected by group and
condition for the second session:
G1C1 G2C1 G3C1
G1C2 G2C2 G3C2
Simulations
Working on accuracies. Some problems with methods include mixing
strategy use with accuracy - with the natural consequence that the
bin counts differ. accuracy data
chstwob, data not normalized
chsone, data normalized
chstowb, data normalized
Note that empirical 'Correct' values are derived by multiplying the
simulated accuracy by the number observed correct. This means that
'Predicted Observed' will be 0 if either the model predicts no use OR
there were none observed correct in the empirical data.
---
352/384-Sample Strategy Chi-Squares
A similar set of data are produced for strategies:
chstwob, data not normalized
chsone, data normalized
chstowb, data normalized
These three sets are based on 352/384 samples (11/12 children * 8
trials * 4 instructions/trial). They are a subset of the original
data generated by the simulations (1408/1536 samples).
---
32-Sample (averaged) Strategy Chi-Squares
Since these are problematic, a set using 32-sample sets is
produced:
chstwob, data not normalized
chsone, data normalized
chstowb, data normalized data
These are the original data (1408/1536 samples) divided by 44/48 (4
ins * 11/12 children) to produce sums of 32.0 (as in total trials per
child). It is this set of data (the chstwob data) that was used to
generate the strategy use patterns as seen in the grant and
publications.
---
(1997-09-23/24 mfv)
Chi-squares of 'Correct' Recall, 'Observed' (empirical) vs.
'Predicted Observed' (model)
* Per Bray meeting, Values from file are divided by 12 so that maximum
possible values (only one strategy used, all correctly recalled) at 88/96
(11/12 subjects). This is similar to the strategy case (which sum to 32)
but we have three categories here (O/R/T).
[*.averaged_accuracy_non_normalized]
G1C1 G2C1 G3C1
G1C2 G2C2 G3C2
Same, but using chsone() test with normalized data.
[*.averaged_accuracy_normalized]
G1C1 G2C1 G3C1
G1C2 G2C2 G3C2
* (1997-10-02 mfv) Update 'echs3'output format to separate user and
machine-readable portions and subset and experiment
prefix/sufix in default output file name.