COR 1022: Cognition - Letter/Category Fluency Module Analysis Data Error Originated in "gcogflu.do" from the 2004 Collection Wave, and WLS Staff's Solution as Implemented in the Official Release of 2011 Data Noah C. Pearce 04/26/2013 i. Path for "gcogflu.do" = /project/hauser/wls/Variable_Creation_2003_2008/Grad/step2_commandfiles/ ii. Path for "jcogflu.do" = /project/hauser/wls/Variable_Creation_2010/Grad/step2_commandfiles/ /****************************************************************************************************/ /*****Description of the problem as it exists in GCOGFLU / CCOGFLU from the 2004 collection wave*****/ /****************************************************************************************************/ WLS staff have discovered several inconsistencies in the analysis data contained within this module, most of them involving the classification of cases as "refusals" (see GI203REC and GI303REC) or "not ascertained" (see GI202REC and GI302REC) across various measures. The following COR is specific to the graduate data, but the same problems and solution are applicable to sibling data as well. Specific issues discovered are as follows: 1. The analysis variables that provide detailed codes for refusals (GI203REC / GI303REC) and "not ascertained" cases (GI202REC / GI302REC) are not mutually exclusive. /* comment: this shows the 33 letter fluency cases coded as not ascertained that also had a detailed refusal code */ tab gi202rec if gi203rec > 0 detailed code for letter fluency not | ascertained | Freq. Percent Cum. ----------------------------------------+----------------------------------- inap (hi211rec!=-4) | 253 88.46 88.46 missing or too poor audio recording | 31 10.84 99.30 severe task misunderstanding or confusi | 1 0.35 99.65 misheard letter, not properly corrected | 1 0.35 100.00 ----------------------------------------+----------------------------------- Total | 286 100.00 /* comment: this shows the 23 category fluency cases coded as not ascertained that also had a detailed refusal code */ tab gi302rec if gi303rec > 0 detailed code for category fluency not | ascertained | Freq. Percent Cum. ----------------------------------------+----------------------------------- inap (hi311rec!=-4) | 117 83.57 83.57 missing or too poor audio recording | 23 16.43 100.00 ----------------------------------------+----------------------------------- Total | 140 100.00 As a consequence of this, the number of refusals and "not ascertained" cases on subsequent analysis variables does not always match the frequencies apparent in the detailed code variables. In the distribution of GI210REC, for example, cases with both detailed refusal and "not ascertained" codes are predominantly lumped into the "not ascertained" category, resulting in there being 41 fewer cases coded as refusals than appear in GI203REC. /* comment: here, the 449 cases coded as "not ascertained" match up with all cases assigned detailed codes at GI202REC; the 245 refusal cases, however, are 41 short of all cases assigned detailed codes at GI203REC */ tab gi210rec if gi210rec < 1 letter fluency: | scored words | named | Freq. Percent Cum. ------------------+----------------------------------- not ascertained | 449 19.54 19.54 refused | 245 10.66 30.20 inap (gi201re==2) | 1,600 69.63 99.83 0 | 4 0.17 100.00 ------------------+----------------------------------- Total | 2,298 100.00 /* comment: here, the 245 cases coded as "not ascertained" match up with all cases assigned detailed codes at GI302REC; the 99 refusal cases, however, are 41 short of all cases assigned detailed codes at GI303REC */ tab gi310rec if gi310rec < 1 category fluency, | quartile : scored | words named | Freq. Percent Cum. ------------------+----------------------------------- not ascertained | 245 6.12 6.12 refused | 99 2.47 8.59 inap (gi301re==2) | 3,658 91.34 99.93 0 | 3 0.07 100.00 ------------------+----------------------------------- Total | 4,005 100.00 Complicating matters further, some of the cases that were assigned detailed refusal codes were not even in the sub-samples randomly selected to complete the fluency tasks; such cases should have been coded as inappropriate on GI203REC and GI303REC. /* comment: 8 cases that were assigned detailed letter fluency refusal codes were not even eligible for the task */ tab gi201re if gi203rec > 0 sample flag | for letter | fluency | task | Freq. Percent Cum. ------------+----------------------------------- 0 | 8 2.80 2.80 1 | 278 97.20 100.00 ------------+----------------------------------- Total | 286 100.00 /* comment: 18 cases that were assigned detailed category fluency refusal codes were not even eligible for the task */ tab gi301re if gi303rec > 0 sample flag | for | category | fluency | task | Freq. Percent Cum. ------------+----------------------------------- 0 | 18 12.86 12.86 1 | 122 87.14 100.00 ------------+----------------------------------- Total | 140 100.00 The inconsistency in refusals between the detailed code variables and subsequent analysis variables summarizing participants' performance on whole cognitive fluency tasks can be entirely attributed to the aforementioned mutual exclusivity problem, as well as the inclusion of cases outside of the Cognition - Letter/Category Module sampling frame. /* comment: of the 286 total cases assigned a detailed letter fluency refusal code, 33 of them also had a code on the detailed "not ascertained" analysis variable and were coded as "not ascertained" on subsequent variables while 8 of them were not even in the relevant sub- sample and were correctly coded as "inappropriate" on subsequent variables */ tab gi210rec if gi202rec > 0 | gi203rec > 0 letter fluency: | scored words | named | Freq. Percent Cum. ------------------+----------------------------------- not ascertained | 449 63.96 63.96 refused | 245 34.90 98.86 inap (gi201re==2) | 8 1.14 100.00 ------------------+----------------------------------- Total | 702 100.00 /* comment: of the 140 total cases assigned a detailed category fluency refusal code, 23 of them also had a code on the detailed "not ascertained" analysis variable and were coded as "not ascertained" on subsequent variables while 18 of them were not even in the relevant sub-sample and were correctly coded as "inappropriate" on subsequent variables */ tab gi310rec if gi302rec > 0 | gi303rec > 0 category fluency, | quartile : scored | words named | Freq. Percent Cum. ------------------+----------------------------------- not ascertained | 245 67.68 67.68 refused | 99 27.35 95.03 inap (gi301re==2) | 18 4.97 100.00 ------------------+----------------------------------- Total | 362 100.00 2. The lack of mutual exclusivity between the detailed refusal and "not ascertained" code analysis variables was not dealt with in a consistent manner for variables summarizing participants' performance on entire cognitive fluency tasks versus those focusing on the first 15-second quartile of the tasks. For variables summarizing participant performance on the tasks in their entirety, cases with overlapping refusal and "not ascertained" detailed codes were coded as "not ascertained". For variables addressing the first 15-second quartile of the tasks, however, cases with overlapping refusal and "not ascertained" detailed codes were coded as refusals. /* comment: on variables about the entire Letter Fluency task, such as GI210REC, cases with both detailed refusal and "not ascertained" codes were coded as "not ascertained" */ tab gi210rec if gi202rec > 0 & gi203rec > 0 letter fluency: | scored words | named | Freq. Percent Cum. ------------------+----------------------------------- not ascertained | 33 100.00 100.00 ------------------+----------------------------------- Total | 33 100.00 /* comment: on variables about quartile #1 of the Letter Fluency task, such as GI220REC, cases with both detailed refusal and "not ascertained" codes were coded as "refused" */ tab gi220rec if gi202rec > 0 & gi203rec > 0 letter fluency, | quartile 1: | scored words | named | Freq. Percent Cum. ----------------+----------------------------------- refused | 33 100.00 100.00 ----------------+----------------------------------- Total | 33 100.00 /* comment: on variables about the entire Category Fluency task, such as GI310REC, cases with both detailed refusal and "not ascertained" codes were coded as "not ascertained" */ tab gi310rec if gi302rec > 0 & gi303rec > 0 category fluency, | quartile : scored | words named | Freq. Percent Cum. ------------------+----------------------------------- not ascertained | 23 100.00 100.00 ------------------+----------------------------------- Total | 23 100.00 /* comment: on variables about quartile #1 of the Category Fluency task, such as GI320REC, cases with both detailed refusal and "not ascertained" codes were coded as "refused" */ tab gi320rec if gi302rec > 0 & gi303rec > 0 category | fluency, | quartile 1: | scored words | named | Freq. Percent Cum. ----------------+----------------------------------- refused | 23 100.00 100.00 ----------------+----------------------------------- Total | 23 100.00 3. Refusal and "not ascertained" codes were also inconsistently applied across 15-second quartiles for both cognitive fluency tasks. While analysis variables addressing quartile #1 for both tasks generally adhere to the detailed refusal and "not ascertained" code variables, those addressing quartiles #2-4 have far less cases coded as refusals and "not ascertained," albeit consistent numbers of them. /* comment: example from quartile #2 for the Letter Fluency task; note that cases coded as "not ascertained" have decreased from 449 to 38 while cases coded as refusals have decreased from 245 to 115. The number of cases coded as refusals or "not ascertained" remains constant across variables addressing quartiles #2-4. */ tab gi228rec if gi228rec < 1 letter fluency, | quartile 2: raw | number of words | said (scored | and | disqualified) | Freq. Percent Cum. ----------------+----------------------------------- not ascertained | 38 1.43 1.43 refused | 115 4.33 5.76 inap | 1,600 60.20 65.95 0 | 905 34.05 100.00 ----------------+----------------------------------- Total | 2,658 100.00 /* comment: example from quartile #2 for the Category Fluency task; note that cases coded as "not ascertained" have decreased from 245 to 17 while cases coded as refusals have decreased from 99 to 40. The number of cases coded as refusals or "not ascertained" remains constant across variables addressing quartiles #2-4. */ tab gi328rec if gi328rec < 1 category | fluency, | quartile 2: raw | number of words | said (scored | and | disqualified) | Freq. Percent Cum. ----------------+----------------------------------- not ascertained | 17 0.42 0.42 refused | 40 0.99 1.41 inap | 3,658 90.52 91.93 0 | 326 8.07 100.00 ----------------+----------------------------------- Total | 4,041 100.00 Making matters worse, cases that were coded as refusals or "not ascertained" in quartile #1 (as well as analysis variables summarizing whole cognitive fluency tasks) have valid data in quartiles #2-4. While most of the affected cases have values of "0" on variables for quartiles #2-4, signifying that no words were volunteered by the participant for the relevant cognitive fluency task, a handful of cases have positive values. /* comment: the following two tabulations show cases with detailed refusal and/or "not ascertained" codes that also have positive, nonzero values on variables addressing quartile #2 of the cognitive fluency tasks. In total, 68 cases said a word (whether "scorable" or not) during quartile #2 of the Letter Fluency task while 13 said a word (whether "scorable" or not) during quartile #2 of the Category Fluency task. */ tab gi228rec if gi202rec > 0 | gi203rec > 0 letter fluency, | quartile 2: raw | number of words | said (scored | and | disqualified) | Freq. Percent Cum. ----------------+----------------------------------- not ascertained | 38 5.41 5.41 refused | 115 16.38 21.79 inap | 8 1.14 22.93 0 | 473 67.38 90.31 1 | 17 2.42 92.74 2 | 17 2.42 95.16 3 | 16 2.28 97.44 4 | 10 1.42 98.86 5 | 5 0.71 99.57 6 | 3 0.43 100.00 ----------------+----------------------------------- Total | 702 100.00 tab gi328rec if gi302rec > 0 | gi303rec > 0 category | fluency, | quartile 2: raw | number of words | said (scored | and | disqualified) | Freq. Percent Cum. ----------------+----------------------------------- not ascertained | 17 4.70 4.70 refused | 40 11.05 15.75 inap | 18 4.97 20.72 0 | 274 75.69 96.41 1 | 2 0.55 96.96 2 | 2 0.55 97.51 3 | 2 0.55 98.07 4 | 1 0.28 98.34 5 | 3 0.83 99.17 6 | 1 0.28 99.45 8 | 1 0.28 99.72 9 | 1 0.28 100.00 ----------------+----------------------------------- Total | 362 100.00 /* comment: the following two tabulations show cases with detailed refusal and/or "not ascertained" codes that also have positive, nonzero values on variables addressing quartile #3 of the cognitive fluency tasks. In total, 60 cases said a word (whether "scorable" or not) during quartile #3 of the Letter Fluency task while 12 said a word (whether "scorable" or not) during quartile #3 of the Category Fluency task. */ tab gi235rec if gi202rec > 0 | gi203rec > 0 letter fluency, | quartile 3: raw | number of words | said (scored | and | disqualified) | Freq. Percent Cum. ----------------+----------------------------------- not ascertained | 38 5.41 5.41 refused | 115 16.38 21.79 inap | 8 1.14 22.93 0 | 481 68.52 91.45 1 | 22 3.13 94.59 2 | 16 2.28 96.87 3 | 10 1.42 98.29 4 | 9 1.28 99.57 5 | 1 0.14 99.72 6 | 1 0.14 99.86 9 | 1 0.14 100.00 ----------------+----------------------------------- Total | 702 100.00 tab gi335rec if gi302rec > 0 | gi303rec > 0 category | fluency, | quartile 3: raw | number of words | said (scored | and | disqualified) | Freq. Percent Cum. ----------------+----------------------------------- not ascertained | 17 4.70 4.70 refused | 40 11.05 15.75 inap | 18 4.97 20.72 0 | 275 75.97 96.69 1 | 4 1.10 97.79 4 | 1 0.28 98.07 5 | 2 0.55 98.62 6 | 3 0.83 99.45 7 | 1 0.28 99.72 12 | 1 0.28 100.00 ----------------+----------------------------------- Total | 362 100.00 /* comment: the following two tabulations show cases with detailed refusal and/or "not ascertained" codes that also have positive, nonzero values on variables addressing quartile #4 of the cognitive fluency tasks. In total, 60 cases said a word (whether "scorable" or not) during quartile #4 of the Letter Fluency task while 9 said a word (whether "scorable" or not) during quartile #4 of the Category Fluency task. */ tab gi242rec if gi202rec > 0 | gi203rec > 0 letter fluency, | quartile 4: raw | number of words | said (scored | and | disqualified) | Freq. Percent Cum. ----------------+----------------------------------- not ascertained | 38 5.41 5.41 refused | 115 16.38 21.79 inap | 8 1.14 22.93 0 | 481 68.52 91.45 1 | 12 1.71 93.16 2 | 23 3.28 96.44 3 | 12 1.71 98.15 4 | 8 1.14 99.29 5 | 3 0.43 99.72 6 | 1 0.14 99.86 9 | 1 0.14 100.00 ----------------+----------------------------------- Total | 702 100.00 tab gi342rec if gi302rec > 0 | gi303rec > 0 category | fluency, | quartile 4: raw | number of words | said (scored | and | disqualified) | Freq. Percent Cum. ----------------+----------------------------------- not ascertained | 17 4.70 4.70 refused | 40 11.05 15.75 inap | 18 4.97 20.72 0 | 278 76.80 97.51 2 | 2 0.55 98.07 3 | 4 1.10 99.17 4 | 2 0.55 99.72 7 | 1 0.28 100.00 ----------------+----------------------------------- Total | 362 100.00 4. More inconsistency with regards to inclusion of cases with detailed refusal and/or "not ascertained" codes occurs in analysis variables GI205REC and GI305REC, which are supposed to indicate the number of seconds elapsed since the beginning of cognitive fluency tasks for participants who did not receive the full 60 seconds when attempting them. The variables' names, which simply read "time until termination for incomplete cases, in seconds" make it unclear whether cases with detailed refusal and/or "not ascertained" codes are supposed to be included; examining the distributions for cases with detailed codes does not lend itself to any obvious conclusions either. In short, how GI205REC and GI305REC should be interpreted is unclear. /* comment: this shows the representation of cases with detailed refusal and/or "not ascertained" codes on the variable for "time until termination for incomplete cases, in seconds" for the Letter Fluency task. While a majority of cases (540) have been set to "inappropriate," 162 have valid times. */ tab gi205rec if gi202rec > 0 | gi203rec > 0 time until | termination | for | incomplete | cases | Freq. Percent Cum. ------------+----------------------------------- inap | 540 76.92 76.92 0 | 1 0.14 77.07 9 | 1 0.14 77.21 11 | 1 0.14 77.35 12 | 2 0.28 77.64 15 | 3 0.43 78.06 16 | 1 0.14 78.21 17 | 1 0.14 78.35 20 | 1 0.14 78.49 21 | 2 0.28 78.77 22 | 1 0.14 78.92 23 | 2 0.28 79.20 24 | 4 0.57 79.77 25 | 5 0.71 80.48 26 | 3 0.43 80.91 27 | 1 0.14 81.05 28 | 4 0.57 81.62 29 | 3 0.43 82.05 30 | 8 1.14 83.19 31 | 4 0.57 83.76 32 | 7 1.00 84.76 33 | 4 0.57 85.33 34 | 3 0.43 85.75 35 | 3 0.43 86.18 36 | 4 0.57 86.75 37 | 3 0.43 87.18 38 | 4 0.57 87.75 39 | 3 0.43 88.18 40 | 3 0.43 88.60 41 | 4 0.57 89.17 42 | 2 0.28 89.46 43 | 7 1.00 90.46 44 | 3 0.43 90.88 45 | 13 1.85 92.74 46 | 3 0.43 93.16 48 | 3 0.43 93.59 49 | 3 0.43 94.02 50 | 8 1.14 95.16 51 | 5 0.71 95.87 52 | 5 0.71 96.58 53 | 3 0.43 97.01 54 | 4 0.57 97.58 55 | 4 0.57 98.15 56 | 3 0.43 98.58 57 | 5 0.71 99.29 58 | 3 0.43 99.72 59 | 2 0.28 100.00 ------------+----------------------------------- Total | 702 100.00 /* comment: this shows the representation of cases with detailed refusal and/or "not ascertained" codes on the variable for "time until termination for incomplete cases, in seconds" for the Category Fluency task. While a majority of cases (303) have been set to "inappropriate," 59 have valid times. */ tab gi305rec if gi302rec > 0 | gi303rec > 0 time until | termination | for | incomplete | cases | Freq. Percent Cum. ------------+----------------------------------- inap | 303 83.70 83.70 7 | 1 0.28 83.98 15 | 1 0.28 84.25 16 | 1 0.28 84.53 20 | 1 0.28 84.81 21 | 2 0.55 85.36 24 | 1 0.28 85.64 25 | 1 0.28 85.91 26 | 2 0.55 86.46 27 | 3 0.83 87.29 29 | 1 0.28 87.57 30 | 6 1.66 89.23 31 | 1 0.28 89.50 32 | 1 0.28 89.78 33 | 1 0.28 90.06 34 | 1 0.28 90.33 38 | 1 0.28 90.61 40 | 3 0.83 91.44 41 | 3 0.83 92.27 42 | 1 0.28 92.54 44 | 1 0.28 92.82 45 | 3 0.83 93.65 47 | 2 0.55 94.20 48 | 1 0.28 94.48 49 | 1 0.28 94.75 50 | 2 0.55 95.30 51 | 2 0.55 95.86 52 | 2 0.55 96.41 53 | 3 0.83 97.24 54 | 1 0.28 97.51 55 | 1 0.28 97.79 56 | 4 1.10 98.90 57 | 3 0.83 99.72 59 | 1 0.28 100.00 ------------+----------------------------------- Total | 362 100.00 5. A final problem with the module involves participants who refused when asked if their entire CATI interview could be recorded and continued to refuse when asked for special permission to record just the Letter and Category Fluency tasks. Such cases were given detailed refusal codes instead of detailed "not ascertained" codes (see the tabulations below), which is problematic because the fluency tasks were not even administered to participants who did not consent to having their interview recorded in full or part. Since participants who refused audio recording were never even asked about their willingness to be cognitively assessed vis-a-vis the Letter and Category Fluency tasks, they would have been more appropriately treated as "not ascertained" cases. /* comment: this shows that 42 letter fluency cases treated as refusals simply did not permit us to record their interviews */ tab gi203rec detailed code for letter fluency | refused | Freq. Percent Cum. --------------------------------------+----------------------------------- inap (hi211rec!=-3) | 6,979 96.06 96.06 refusal before task began | 111 1.53 97.59 refusal during administration of task | 133 1.83 99.42 refused audio recording of survey | 42 0.58 100.00 --------------------------------------+----------------------------------- Total | 7,265 100.00 /* comment: this shows that 42 category fluency cases treated as refusals simply did not permit us to record their interviews */ tab gi303rec detailed code for category fluency | refused | Freq. Percent Cum. --------------------------------------+----------------------------------- inap (hi311rec!=-3) | 7,125 98.07 98.07 refusal before task began | 54 0.74 98.82 refusal during administration of task | 44 0.61 99.42 refused audio recording of survey | 42 0.58 100.00 --------------------------------------+----------------------------------- Total | 7,265 100.00 /****************************************************************************************************/ /*****How problems were resolved in the 2011 fluency data, and how users can "fix" the 2004 data*****/ /****************************************************************************************************/ The previously highlighted issues were resolved prior to our official release of the 2011 cognitive fluency data (JCOGFLU / KCOGFLU). At this time, however, WLS staff has no intentions of similarly resolving the inconsistencies as they exist in the analysis data for the 2004 collection wave. Users of the 2004 fluency data can align their analysis variables with 2011 conventions by replicating the following procedure, which was used to resolve all identified issues prior to releasing the 2011 fluency data: 1. Cases coded as "3" (refused audio recording of survey) on GI203REC and GI303REC should be assigned values of "10" on GI202REC and GI302REC; this will signify a new detailed "not ascertained" code: "refused audio recording of module, fluency task not administered." Then, such cases should be recoded to values of "-2" on GI203REC and GI303REC, signifying "inappropriate." Here's some sample SAS code: /* for the Letter Fluency task */ if gi203rec = 3 then do; gi202rec = 10; gi203rec = -2; end; /* for the Category Fluency task */ if gi303rec = 3 then do; gi302rec = 10; gi303rec = -2; end; 2. Cases with both detailed refusal and "not ascertained" codes should be reconciled such that GI202REC/GI302REC and GI203REC/GI303REC are mutually exclusive. Do this by giving refusals precedence over "not ascertained"; that is, recode GI202REC/GI302REC to "-2" for cases with valid data on both GI202REC/GI302REC and GI203REC/GI303REC. Here's some sample SAS code: /* for the Letter Fluency task */ if gi202rec gt 0 and gi203rec gt 0 then do; gi202rec = -2; end; /* for the Category Fluency task */ if gi302rec gt 0 and gi303rec gt 0 then do; gi302rec = -2; end; 3. Account for the newly reconciled detailed refusal and "not ascertained" codes at most of the analysis variables for both cognitive fluency tasks (all of them except for GI201RE/GI301RE, GI204REC/GI304REC, and GI206RE/GI306RE). This should be accomplished as follows: /* for the Letter Fluency task */ if 0 < gi202rec, set the following analysis variables to the indicated value: gi205rec = -2 gi210rec = -4 gi211rec = -4 gi212rec = -4 gi213rec = -4 gi214rec = -4 gi215rec = -4 gi216rec = -4 gi220rec = -4 gi221rec = -4 gi222rec = -4 gi223rec = -4 gi224rec = -4 gi225rec = -4 gi226rec = -4 gi227rec = -4 gi228rec = -4 gi229rec = -4 gi230rec = -4 gi231rec = -4 gi232rec = -4 gi233rec = -4 gi234rec = -4 gi235rec = -4 gi236rec = -4 gi237rec = -4 gi238rec = -4 gi239rec = -4 gi240rec = -4 gi241rec = -4 gi242rec = -4 gi243rec = -4 gi244rec = -4 gi245rec = -4 gi246rec = -4 gi247rec = -4 else if 0 < gi203rec, set the following analysis variables to the indicated value: gi205rec = -2 gi210rec = -3 gi211rec = -3 gi212rec = -3 gi213rec = -3 gi214rec = -3 gi215rec = -3 gi216rec = -3 gi220rec = -3 gi221rec = -3 gi222rec = -3 gi223rec = -3 gi224rec = -3 gi225rec = -3 gi226rec = -3 gi227rec = -3 gi228rec = -3 gi229rec = -3 gi230rec = -3 gi231rec = -3 gi232rec = -3 gi233rec = -3 gi234rec = -3 gi235rec = -3 gi236rec = -3 gi237rec = -3 gi238rec = -3 gi239rec = -3 gi240rec = -3 gi241rec = -3 gi242rec = -3 gi243rec = -3 gi244rec = -3 gi245rec = -3 gi246rec = -3 gi247rec = -3 /* for the Category Fluency task */ if 0 < gi302rec, set the following analysis variables to the indicated value: gi305rec = -2 gi310rec = -4 gi311rec = -4 gi312rec = -4 gi314rec = -4 gi315rec = -4 gi316rec = -4 gi320rec = -4 gi321rec = -4 gi322rec = -4 gi324rec = -4 gi325rec = -4 gi326rec = -4 gi327rec = -4 gi328rec = -4 gi329rec = -4 gi331rec = -4 gi332rec = -4 gi333rec = -4 gi334rec = -4 gi335rec = -4 gi336rec = -4 gi338rec = -4 gi339rec = -4 gi340rec = -4 gi341rec = -4 gi342rec = -4 gi343rec = -4 gi345rec = -4 gi346rec = -4 gi347rec = -4 else if 0 < gi303rec, set the following analysis variables to the indicated value: gi305rec = -2 gi310rec = -3 gi311rec = -3 gi312rec = -3 gi314rec = -3 gi315rec = -3 gi316rec = -3 gi320rec = -3 gi321rec = -3 gi322rec = -3 gi324rec = -3 gi325rec = -3 gi326rec = -3 gi327rec = -3 gi328rec = -3 gi329rec = -3 gi331rec = -3 gi332rec = -3 gi333rec = -3 gi334rec = -3 gi335rec = -3 gi336rec = -3 gi338rec = -3 gi339rec = -3 gi340rec = -3 gi341rec = -3 gi342rec = -3 gi343rec = -3 gi345rec = -3 gi346rec = -3 gi347rec = -3 Note that this step will resolve all of the identified inconsistencies in refusal and "not ascertained" codes between analysis variables summarizing whole cognitive fluency tasks, those addressing quartile #1, and those addressing quartiles #2-4. The step also has implications on how GI205REC/GI305REC ("time until termination for incomplete cases, in seconds") should be interpreted; these analysis variables must now be interpreted as providing the number of seconds participants who did not use the full 60 seconds allotted to them attempted a task for. That is, an "incomplete case" now signifies a participant who did not use the full 60 seconds allotted to them on an appropriately administered cognitive fluency task that they did not refuse (i.e. they have no detailed refusal or "not ascertained" codes and simply quit the task short of 60 seconds). 4. To clear up any detailed refusal or "not ascertained" codes that were assigned to cases not even in the selected sub-samples for a cognitive fluency task, use the sampling flag analysis variables (GI201RE/GI301RE) to set all other variables for a fluency task to "-2". Here's some sample SAS code: /* for the Letter Fluency task */ if gi201re ne 1 then do; gi202rec = -2; gi203rec = -2; gi204rec = -2; gi205rec = -2; gi206re = -2; gi210rec = -2; gi211rec = -2; gi212rec = -2; gi213rec = -2; gi214rec = -2; gi215rec = -2; gi216rec = -2; gi220rec = -2; gi221rec = -2; gi222rec = -2; gi223rec = -2; gi224rec = -2; gi225rec = -2; gi226rec = -2; gi227rec = -2; gi228rec = -2; gi229rec = -2; gi230rec = -2; gi231rec = -2; gi232rec = -2; gi233rec = -2; gi234rec = -2; gi235rec = -2; gi236rec = -2; gi237rec = -2; gi238rec = -2; gi239rec = -2; gi240rec = -2; gi241rec = -2; gi242rec = -2; gi243rec = -2; gi244rec = -2; gi245rec = -2; gi246rec = -2; gi247rec = -2; end; /* for the Category Fluency task */ if gi301re ne 1 then do; gi302rec = -2; gi303rec = -2; gi304rec = -2; gi305rec = -2; gi306re = -2; gi310rec = -2; gi311rec = -2; gi312rec = -2; gi314rec = -2; gi315rec = -2; gi316rec = -2; gi320rec = -2; gi321rec = -2; gi322rec = -2; gi324rec = -2; gi325rec = -2; gi326rec = -2; gi327rec = -2; gi328rec = -2; gi329rec = -2; gi331rec = -2; gi332rec = -2; gi333rec = -2; gi334rec = -2; gi335rec = -2; gi336rec = -2; gi338rec = -2; gi339rec = -2; gi340rec = -2; gi341rec = -2; gi342rec = -2; gi343rec = -2; gi345rec = -2; gi346rec = -2; gi347rec = -2; end; If the above procedure was implemented in order, all of the previously identified inconsistencies in cognitive fluency analysis data should now be resolved.