Hi Jingwei, It looks like the 15 and 14.96 scores for Mars_Final (Contrast sensitivity) were brought over from the raw data for that measure that we get from the UPenn Computerized Behavioral Battery. Likely there was some strange answer entered by the subject that led to these outliers, but we can't know what happened definitively. I would exclude these 2 subjects from your analysis on that measure.
We try to define the codes for "unknown" in the HCP Data Dictionary for measures where "unknown" is a valid answer that a subject gave (e.g. "Correction" for Eyeglass correction, if subject didn't know their eyeglass lens correction in diopters, they answer 999 for unknown). Missing data for a subject is generally just left blank in the spreadsheet. We can check our raw data in particular cases to verify, but in most cases where the data is missing the data was never collected (because the subject refused, or skipped an item or test, left the study visit early, etc.). Many of the tests were self-administered so subjects may score lower because they chose to not complete the test. For example, there are 4 such subjects with 0 scores for the SCPT_SEN (Short Penn Continuous Performance Test Sensitivity) which is due to 0 true positive answers on the test. The abnormally low score for that measure could indicate that these subjects decided not to complete the test. I would suggest that if you see far outliers, you are welcome to ask us directly about them, but be prepared to take outlier subjects out of your analysis for the particular measure. Best, Jenn Jennifer Elam, Ph.D. Scientific Outreach, Human Connectome Project Washington University School of Medicine Department of Neuroscience, Box 8108 660 South Euclid Avenue St. Louis, MO 63110 314-362-9387<tel:314-362-9387> e...@wustl.edu<mailto:e...@wustl.edu> www.humanconnectome.org<http://www.humanconnectome.org/> ________________________________ From: Jingwei Li <jingweili.sjtu....@gmail.com> Sent: Thursday, September 14, 2017 10:25:04 PM To: Elam, Jennifer Cc: hcp-users@humanconnectome.org; Hodge, Michael Subject: Re: [HCP-Users] HCP behavioral data "Noise_Comp" with -99 Hi Jenn, Except for Noise_Comp, we also found a weird number -15 (and -14.96) for Mar_Final because all the other values are around 1-2 in this behavior. So we are eager to know what is the coding system to mark "missing/erroneous data" for each behavior. Here is the list of behaviors we are using: PicSeq_Unadj CardSort_Unadj Flanker_Unadj PMAT24_A_CR PMAT24_A_SI ReadEng_Unadj PicVocab_Unadj ProcSpeed_Unadj DDisc_AUC_40K VSPLOT_TC SCPT_SEN * SCPT_SPEC IWRD_TOT ListSort_Unadj MMSE_Score PSQI_Score Endurance_Unadj GaitSpeed_Comp Dexterity_Unadj Strength_Unadj Noise_Comp Odor_Unadj PainInterf_Tscore Taste_Unadj Mars_Final Emotion_Task_Face_Acc Language_Task_Math_Avg_Difficulty_Level Language_Task_Story_Avg_Difficulty_Level Relational_Task_Acc Social_Task_Perc_Random Social_Task_Perc_TOM WM_Task_Acc NEOFAC_A NEOFAC_O NEOFAC_C NEOFAC_N NEOFAC_E ER40_CR ER40ANG ER40FEAR ER40HAP ER40NOE ER40SAD AngAffect_Unadj AngHostil_Unadj AngAggr_Unadj FearAffect_Unadj FearSomat_Unadj Sadness_Unadj LifeSatisf_Unadj MeanPurp_Unadj PosAffect_Unadj Friendship_Unadj Loneliness_Unadj PercHostil_Unadj PercReject_Unadj EmotSupp_Unadj InstruSupp_Unadj PercStress_Unadj SelfEff_Unadj Sorry it is a long list. I am so appreciated with your help. Thanks, Jingwei On Fri, Sep 15, 2017 at 8:47 AM, Jingwei Li <jingweili.sjtu....@gmail.com<mailto:jingweili.sjtu....@gmail.com>> wrote: Hi Jennifer, thanks. --Jingwei 在 2017年9月15日,上午1:45,Elam, Jennifer <e...@wustl.edu<mailto:e...@wustl.edu>> 写道: Hi Jingwei, The issue you found is limited to the "Noise_Comp" variable, which is the NIH Toolbox Words in Noise (WIN) Computed score. NIH Toolbox changed the WIN test from v1 to v2 during the HCP data collection and only the v1 scores have been correctly released in ConnectomeDB. Those subjects that have -99, and many of of those with no score were tested with WIN v2. Originally, our internal database was keeping the v1 and v2 scores separate so that we could determine comparability b/w v1 and v2. NIH Toolbox has verified that the v1 and v2 scoring has been normed to be directly comparable, so we need to combine the v1/v2 scores into a single variable so all scores are released. Therefore, we need to fix the released scores for the "Noise_Comp" variable -- we will document the issue on the HCP Issues and Planned Fixes wiki page and email the list when we release the fix. Best, Jenn Jennifer Elam, Ph.D. Scientific Outreach, Human Connectome Project Washington University School of Medicine Department of Neuroscience, Box 8108 660 South Euclid Avenue St. Louis, MO 63110 314-362-9387<tel:314-362-9387> e...@wustl.edu<mailto:e...@wustl.edu> www.humanconnectome.org<http://www.humanconnectome.org/> ________________________________ From: hcp-users-boun...@humanconnectome.org<mailto:hcp-users-boun...@humanconnectome.org> <hcp-users-boun...@humanconnectome.org<mailto:hcp-users-boun...@humanconnectome.org>> on behalf of Jingwei Li <jingweili.sjtu....@gmail.com<mailto:jingweili.sjtu....@gmail.com>> Sent: Thursday, September 14, 2017 6:36:41 AM To: hcp-users@humanconnectome.org<mailto:hcp-users@humanconnectome.org> Subject: [HCP-Users] HCP behavioral data "Noise_Comp" with -99 Dear HCP experts, We are using the HCP behavioral data and noticed that for the behavior "Noise_Comp", there are a lot of subjects with a value of -99. We want to ask whether this number means erroneous data. Are those subjects usable for this behavior? If -99 means error and is not usable, is there any other behavior that has similiar issue that we need to do a postprocessing? Thanks, Jingwei _______________________________________________ HCP-Users mailing list HCP-Users@humanconnectome.org<mailto:HCP-Users@humanconnectome.org> http://lists.humanconnectome.org/mailman/listinfo/hcp-users _______________________________________________ HCP-Users mailing list HCP-Users@humanconnectome.org http://lists.humanconnectome.org/mailman/listinfo/hcp-users