{ "id": 10815, "title": "Understanding the data", "url": "/competitions/seizure-prediction/discussion/10815", "author_name": "", "post_date": "2014-11-01T12:21:08.530Z", "votes": null, "comment_count": 9, "views": 2058, "content": "

It's not clear to me what the values for the electrodes represent. This is what Kaggle says about the data: "Intracranial EEG was recorded from dogs with naturally occurring epilepsy using an ambulatory monitoring system. EEG was sampled from 16 electrodes at 400 Hz, and recorded voltages were referenced to the group average." 

\n

After some research, I think I understand what sampled at 400 Hz means, but I am still unclear about the recorded voltages referenced to the group averages.

\n

I would like to know what is meant by a value of -26 for eletrode NVC1202_32_002_Ecog_c002 for Dog_1?

\n

Any guidance or references to resources would be greatly appreciated.

\n

Thanks.

", "messages": [ { "id": "57161", "postDate": "11/01/2014 12:21:08", "content": "

It's not clear to me what the values for the electrodes represent. This is what Kaggle says about the data: "Intracranial EEG was recorded from dogs with naturally occurring epilepsy using an ambulatory monitoring system. EEG was sampled from 16 electrodes at 400 Hz, and recorded voltages were referenced to the group average." 

\n

After some research, I think I understand what sampled at 400 Hz means, but I am still unclear about the recorded voltages referenced to the group averages.

\n

I would like to know what is meant by a value of -26 for eletrode NVC1202_32_002_Ecog_c002 for Dog_1?

\n

Any guidance or references to resources would be greatly appreciated.

\n

Thanks.

", "rawMarkdown": "", "votes": null }, { "id": "57166", "postDate": "11/01/2014 17:03:15", "content": "

Hi,

\n

When we record the EEG there should be reference. Typically the referential voltage will be the EAR electrode for human scalp and depth EEG. Sometimes for artifact correction we average the EEG data across all the channels and subtracts the averaged data from individual data.

\n

For example if X1, X2, X3 are the original data recorded. Then with group averaged reference the new voltage will be

\n

Xn1 = (X1 - (X1+X2+X3)/3)

\n

Xn2 =  (X2 - (X1+X2+X3)/3) etc

\n\n

I am not sure about what you meant by -26. Hope I am clear

\n\n\n\n

Ho

", "rawMarkdown": "", "votes": null }, { "id": "57172", "postDate": "11/01/2014 18:50:05", "content": "

Hi Balu,

\n

Thanks for your reply.

\n

The -26 that I was referring to is the first value for the column NVC1202_32_002_Ecog_c002 (it's a column for me because I transposed it).

\n

Is that -26 value the (voltage - the average(voltage) for NVC1202_32_002_Ecog_c002)?

\n

thanks,

\n

C

", "rawMarkdown": "", "votes": null }, { "id": "57177", "postDate": "11/02/2014 05:17:30", "content": "

[quote=Balu Krishnan;57166]

\n

Hi,

\n

When we record the EEG there should be reference. Typically the referential voltage will be the EAR electrode for human scalp and depth EEG. Sometimes for artifact correction we average the EEG data across all the channels and subtracts the averaged data from individual data.

\n

For example if X1, X2, X3 are the original data recorded. Then with group averaged reference the new voltage will be

\n

Xn1 = (X1 - (X1+X2+X3)/3)

\n

Xn2 =  (X2 - (X1+X2+X3)/3) etc

\n

I am not sure about what you meant by -26. Hope I am clear

\n

Ho

\n

[/quote]

\n\n

Is the mean on each file or mean on all types of files of one subject?

", "rawMarkdown": "", "votes": null }, { "id": "57197", "postDate": "11/02/2014 16:44:48", "content": "

It is the mean on each file (i.e. every 10 minutes of data segment).. Just to make sure I explain it clearly

\n

Suppose there are only 4 channels in Dog_1 and the time series for channel is given as

\n

X1: X1(1), X2(2)..............X1(N)

\n

X2: X2(1), X2(2) .............X2(N)  and so forth

\n

then the average reference will be

\n

XAVG : (X1(1)+X2(1)..X4(1))/4 ,    (X1(2)+X2(2)...X4(2))/4 ....... and so forth.

", "rawMarkdown": "", "votes": null }, { "id": "57203", "postDate": "11/02/2014 18:10:03", "content": "

[quote=Balu Krishnan;57197]

\n

It is the mean on each file (i.e. every 10 minutes of data segment).. Just to make sure I explain it clearly

\n

Suppose there are only 4 channels in Dog_1 and the time series for channel is given as

\n

X1: X1(1), X2(2)..............X1(N)

\n

X2: X2(1), X2(2) .............X2(N)  and so forth

\n

then the average reference will be

\n

XAVG : (X1(1)+X2(1)..X4(1))/4 ,    (X1(2)+X2(2)...X4(2))/4 ....... and so forth.

\n

[/quote]

\n

From your example, if original data is

\n

X1:60, 60, 60, 60

\n

X2:20, 20, 20, 20

\n

X3:0,   0,    0,    0

\n

X4:0,  40,  0,  40

\n

will become:

\n

X1:40,  30,   40, 30  

\n

X2:0,  -10,    0,   -10

\n

X3:-20, -30, -20 ,-30

\n

X4:-20, 10,  -20  ,10

\n\n

Is it correct?

", "rawMarkdown": "", "votes": null }, { "id": "57204", "postDate": "11/02/2014 18:12:49", "content": "

Yes. You are correct

", "rawMarkdown": "", "votes": null }, { "id": "57206", "postDate": "11/02/2014 18:39:00", "content": "

[quote=Balu Krishnan;57204]

\n

Yes. You are correct

\n

[/quote]

\n\n

Seems after averaging, my DC channel got distorted.

\n

 60,60,60,60 become 40,30,40,30

", "rawMarkdown": "", "votes": null }, { "id": "57209", "postDate": "11/02/2014 19:05:01", "content": "

Hi Balu, Steven:

\n

The explanations and examples from both of you have helped me greatly. Just want to be sure I understand.

\n

A sample of the the channels for Dog_1 are below. And what would be considered a record, in database parlance, is taking the first time interval for each channel such that:

\n

NVC1202_32_002_Ecog_c001 = [60, 60, 60, 60]

\n

NVC1202_32_002_Ecog_c002 = [20, 20, 20, 20]

\n

NVC1202_32_002_Ecog_c003 = [0, 0, 0, 0]

\n

NVC1202_32_002_Ecog_c004 = [0, 40, 0, 40]

\n

The average cut across channels:

\n

( NVC1202_32_002_Ecog_c001[0] + NVC1202_32_002_Ecog_c002[0] + NVC1202_32_002_Ecog_c003[0] + NVC1202_32_002_Ecog_c004[0] ) / 4 = 20

\n

so they become:                                Time Interval 0

\n

NVC1202_32_002_Ecog_c001[0]  =   [40]

\n

NVC1202_32_002_Ecog_c002[0]  =    [0]

\n

NVC1202_32_002_Ecog_c003[0]  =   [-20]

\n

NVC1202_32_002_Ecog_c004[0]  =   [-20]

\n\n

Is that right?

\n

Thanks,

\n

C

", "rawMarkdown": "", "votes": null }, { "id": "57572", "postDate": "11/10/2014 02:31:26", "content": "

1) I cannot reconcile the proposed normalization with data: If a mean is subtracted from entries, then the sum of mean-adjusted numbers should be zero. I have tried different summations (across rows, columns, whole matrix) and none of them is zero!

\n

 2) Any reason all data are integers? I do not have MATLAB and I am just wondering if I have made an error while importing data into python ...

", "rawMarkdown": "", "votes": null } ], "comments": [ { "id": 57166, "author_name": "balukrishnan", "author_url": "", "post_date": "11/01/2014 17:03:15", "content": "

Hi,

\n

When we record the EEG there should be reference. Typically the referential voltage will be the EAR electrode for human scalp and depth EEG. Sometimes for artifact correction we average the EEG data across all the channels and subtracts the averaged data from individual data.

\n

For example if X1, X2, X3 are the original data recorded. Then with group averaged reference the new voltage will be

\n

Xn1 = (X1 - (X1+X2+X3)/3)

\n

Xn2 =  (X2 - (X1+X2+X3)/3) etc

\n\n

I am not sure about what you meant by -26. Hope I am clear

\n\n\n\n

Ho

", "votes": null, "replies": [] }, { "id": 57172, "author_name": "chuzzelwit", "author_url": "", "post_date": "11/01/2014 18:50:05", "content": "

Hi Balu,

\n

Thanks for your reply.

\n

The -26 that I was referring to is the first value for the column NVC1202_32_002_Ecog_c002 (it's a column for me because I transposed it).

\n

Is that -26 value the (voltage - the average(voltage) for NVC1202_32_002_Ecog_c002)?

\n

thanks,

\n

C

", "votes": null, "replies": [] }, { "id": 57177, "author_name": "stevendu", "author_url": "", "post_date": "11/02/2014 05:17:30", "content": "

[quote=Balu Krishnan;57166]

\n

Hi,

\n

When we record the EEG there should be reference. Typically the referential voltage will be the EAR electrode for human scalp and depth EEG. Sometimes for artifact correction we average the EEG data across all the channels and subtracts the averaged data from individual data.

\n

For example if X1, X2, X3 are the original data recorded. Then with group averaged reference the new voltage will be

\n

Xn1 = (X1 - (X1+X2+X3)/3)

\n

Xn2 =  (X2 - (X1+X2+X3)/3) etc

\n

I am not sure about what you meant by -26. Hope I am clear

\n

Ho

\n

[/quote]

\n\n

Is the mean on each file or mean on all types of files of one subject?

", "votes": null, "replies": [] }, { "id": 57197, "author_name": "balukrishnan", "author_url": "", "post_date": "11/02/2014 16:44:48", "content": "

It is the mean on each file (i.e. every 10 minutes of data segment).. Just to make sure I explain it clearly

\n

Suppose there are only 4 channels in Dog_1 and the time series for channel is given as

\n

X1: X1(1), X2(2)..............X1(N)

\n

X2: X2(1), X2(2) .............X2(N)  and so forth

\n

then the average reference will be

\n

XAVG : (X1(1)+X2(1)..X4(1))/4 ,    (X1(2)+X2(2)...X4(2))/4 ....... and so forth.

", "votes": null, "replies": [] }, { "id": 57203, "author_name": "stevendu", "author_url": "", "post_date": "11/02/2014 18:10:03", "content": "

[quote=Balu Krishnan;57197]

\n

It is the mean on each file (i.e. every 10 minutes of data segment).. Just to make sure I explain it clearly

\n

Suppose there are only 4 channels in Dog_1 and the time series for channel is given as

\n

X1: X1(1), X2(2)..............X1(N)

\n

X2: X2(1), X2(2) .............X2(N)  and so forth

\n

then the average reference will be

\n

XAVG : (X1(1)+X2(1)..X4(1))/4 ,    (X1(2)+X2(2)...X4(2))/4 ....... and so forth.

\n

[/quote]

\n

From your example, if original data is

\n

X1:60, 60, 60, 60

\n

X2:20, 20, 20, 20

\n

X3:0,   0,    0,    0

\n

X4:0,  40,  0,  40

\n

will become:

\n

X1:40,  30,   40, 30  

\n

X2:0,  -10,    0,   -10

\n

X3:-20, -30, -20 ,-30

\n

X4:-20, 10,  -20  ,10

\n\n

Is it correct?

", "votes": null, "replies": [] }, { "id": 57204, "author_name": "balukrishnan", "author_url": "", "post_date": "11/02/2014 18:12:49", "content": "

Yes. You are correct

", "votes": null, "replies": [] }, { "id": 57206, "author_name": "stevendu", "author_url": "", "post_date": "11/02/2014 18:39:00", "content": "

[quote=Balu Krishnan;57204]

\n

Yes. You are correct

\n

[/quote]

\n\n

Seems after averaging, my DC channel got distorted.

\n

 60,60,60,60 become 40,30,40,30

", "votes": null, "replies": [] }, { "id": 57209, "author_name": "chuzzelwit", "author_url": "", "post_date": "11/02/2014 19:05:01", "content": "

Hi Balu, Steven:

\n

The explanations and examples from both of you have helped me greatly. Just want to be sure I understand.

\n

A sample of the the channels for Dog_1 are below. And what would be considered a record, in database parlance, is taking the first time interval for each channel such that:

\n

NVC1202_32_002_Ecog_c001 = [60, 60, 60, 60]

\n

NVC1202_32_002_Ecog_c002 = [20, 20, 20, 20]

\n

NVC1202_32_002_Ecog_c003 = [0, 0, 0, 0]

\n

NVC1202_32_002_Ecog_c004 = [0, 40, 0, 40]

\n

The average cut across channels:

\n

( NVC1202_32_002_Ecog_c001[0] + NVC1202_32_002_Ecog_c002[0] + NVC1202_32_002_Ecog_c003[0] + NVC1202_32_002_Ecog_c004[0] ) / 4 = 20

\n

so they become:                                Time Interval 0

\n

NVC1202_32_002_Ecog_c001[0]  =   [40]

\n

NVC1202_32_002_Ecog_c002[0]  =    [0]

\n

NVC1202_32_002_Ecog_c003[0]  =   [-20]

\n

NVC1202_32_002_Ecog_c004[0]  =   [-20]

\n\n

Is that right?

\n

Thanks,

\n

C

", "votes": null, "replies": [] }, { "id": 57572, "author_name": "rezamahani", "author_url": "", "post_date": "11/10/2014 02:31:26", "content": "

1) I cannot reconcile the proposed normalization with data: If a mean is subtracted from entries, then the sum of mean-adjusted numbers should be zero. I have tried different summations (across rows, columns, whole matrix) and none of them is zero!

\n

 2) Any reason all data are integers? I do not have MATLAB and I am just wondering if I have made an error while importing data into python ...

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "57161": "", "57166": "", "57172": "", "57177": "", "57197": "", "57203": "", "57204": "", "57206": "", "57209": "", "57572": "" }, "source": "meta" }