add blending experiment to db

[tjmahr]

Will first need a description of what each of the Eprime fields mean and what the best outcome measures would be for the database.

[tjmahr]

Were there different dialect version of this task?

Update: No.

[tjmahr]

Looks like there are three kinds of trial procedures in the Blending experiment. Not sure what's going on in the third kind of procedure.

Level Counts: 
 Eprime.Level      Running    Procedure freq
            1       Header       Header    1
            2 PracticeList PracticeProc    3
            2    TrialList    TrialProc   11
            2        List1   TrialProc2   11
            1         <NA>         <NA>    1

 Eprime.Level      Running    Procedure
            2 PracticeList PracticeProc
List of 15
 $ Eprime.Level      : num 2
 $ Eprime.LevelName  : chr "PracticeList_1"
 $ Eprime.FrameNumber: chr "2"
 $ Procedure         : chr "PracticeProc"
 $ Running           : chr "PracticeList"
 $ Stimulus1         : chr "cowboy"
 $ Stimulus2         : chr "toothbrush"
 $ Stimulus3         : chr "fireman"
 $ StimPrompt        : chr "tooth_brush"
 $ CorrectResponse   : chr "toothbrush"
 $ Cycle             : chr "1"
 $ Sample            : chr "1"
 $ Correct           : chr "1"
 $ Response          : chr "toothbrush"
 - attr(*, "class")= chr [1:2] "EprimeFrame" "list"

 Eprime.Level   Running Procedure
            2 TrialList TrialProc
List of 15
 $ Eprime.Level      : num 2
 $ Eprime.LevelName  : chr "TrialList_1"
 $ Eprime.FrameNumber: chr "5"
 $ Procedure         : chr "TrialProc"
 $ Running           : chr "TrialList"
 $ Stimulus1         : chr "pancake"
 $ Stimulus2         : chr "seashell"
 $ Stimulus3         : chr "mailman"
 $ StimPrompt        : chr "pan_cake"
 $ CorrectResponse   : chr "pancake"
 $ Correct           : chr "1"
 $ Response          : chr "pancake"
 $ Cycle             : chr "1"
 $ Sample            : chr "1"
 - attr(*, "class")= chr [1:2] "EprimeFrame" "list"

 Eprime.Level Running  Procedure
            2   List1 TrialProc2
List of 11
 $ Eprime.Level      : num 2
 $ Eprime.LevelName  : chr "List1_1"
 $ Eprime.FrameNumber: chr "16"
 $ Procedure         : chr "TrialProc2"
 $ Running           : chr "List1"
 $ StimPrompt        : chr "sand_box"
 $ Correct           : chr "0"
 $ Cycle             : chr "1"
 $ Sample            : chr "1"
 $ LadderPosition    : chr "Position1"
 - attr(*, "class")= chr [1:2] "EprimeFrame" "list"

[tjmahr]

Notes on the experiment from ME.

All children received the blending experiment in mainstream American English.

The blending experiment has a total of three procedures.The first procedure (PracticeProc) corresponds to the training or practice trials consisting of visual + auditory stimulus trials. The second procedure (TrialProc) corresponds to the visual + auditory stimulus trials. Lastly, the third procedure (TrialProc2) corresponds to the auditory only stimulus trials.

Procedure 1 (PracticeProc) - 3 trials

1. Children are presented with images that are labeled by the computer (Stimulus1, Stimulus2, Stimulus3).
2. Following the images, a stimulus prompt is played "What do you get when you say cow boy together? (StimPrompt)
3. A field corresponding to the correct response -- image to be selected (Correct Response)
4. The response provided by the child ?? (Response)

Procedure 2 (TrialProc) - same as PracticeProc with 11 trials

Procedure 3 (TrialProc2) - different from the above procedures (auditory stim only)

1. Children presented with the verbal prompt, an auditory only prompt "What do you get when you say sand box together?" (StimPrompt)
2. I can't open the edat files to see what the columns look like yet, [so] I'm going to make a wild guess ... but to record the child's verbal responses, examiners used a keyboard response where "C" corresponded to correct responses and and "I" corresponded to incorrect responses. I'm not quite sure how E-prime decided to treat these responses.
3. A visual reinforcement was used to maintain children's cooperation (LadderPosition).

Outcome measures

This is just a rough sketch of what I'm thinking because not only does this experiment have two conditions (VA and A only) but it also has 3 levels of item complexity (compound word blending, syllable blending and phoneme blending). An additional layer of complexity [... is that] not all children receive all stimulus items. If 5 items are missed in a row the experiment either continues to the second condition or terminates if they were able to advance to the second condition. If the non-administered trials are just coded as incorrect then this computation is easy and of minimal concern and then you can can ignore this comment completely.

• A set of global summary measures
  1. % correct (trials correct in VA +A)
  2. % correct in VA only condition
  3. % correct in A only condition
• A set of of sub-level summary measures:
  1. [% correct by 2 conditions by 3 complexity types]

[tjmahr]

Looks like the practice section could be repeated, so don't assume n = 3 trials in practice section.

      TrialType NumTrials NumFilesWithThatManyTrials
          <chr>     <int>                      <int>
1         List1         5                          7
2         List1         6                          2
3         List1         7                          1
4         List1         8                          1
5         List1         9                          4
6         List1        10                          2
7         List1        11                         48
8  PracticeList         3                         59
9  PracticeList         6                          7
10    TrialList         1                          1
11    TrialList         5                          1
12    TrialList         6                          1
13    TrialList         9                          1
14    TrialList        11                         62

[tjmahr]

Pushed data from 1,670 trials in 66 administrations to the database. Looks like only TimePoint2 participants received the experiment.

Because we aborted experiments after 5 incorrects, I added an Administered (0/1) field for whether a trial was administered, so that we can differentiate implicit/explicit incorrect responses in Correct (NA/0/1).

Next, I need to archive Blending_673L in which experiment was halted midway through first test trial, and check that their Blending data isn't in the database.

And I need to write queries to summarize the raw data and add summary stats to the q_Scores_TimePoint2 table.