help

help >

**RE: Fischer transformed coefficients (RRC matrix) values >1**Feb 26, 2021 05:02 PM | Alfonso Nieto-Castanon -

*Boston University*RE: Fischer transformed coefficients (RRC matrix) values >1

Hi Melissa,

While r values range between -1 and 1, Fisher transformed values (defined as Z = atanh(r), r = tanh(Z)) range between -inf and inf (e.g. a value of Z=1 simply corresponds to r=0.7616, a value of Z=2 corresponds to r=0.9640, Z=3 to r=0.9951, etc.)

Best

Alfonso

While r values range between -1 and 1, Fisher transformed values (defined as Z = atanh(r), r = tanh(Z)) range between -inf and inf (e.g. a value of Z=1 simply corresponds to r=0.7616, a value of Z=2 corresponds to r=0.9640, Z=3 to r=0.9951, etc.)

Best

Alfonso

*Originally posted by Melissa DiFabio:*Hello,

I was hoping to get some clarification on the outputs from CONN's first-level Seed based analysis (SBC). I exported the conditionxxx.mat files for each of my subjects/conditions and pulled the Fischer transformed correlation coefficients (the "Z") in order to run my own statistics in another program. I've noticed that I some few values that are >1; for example, I have values as high as 1.4493 and 1.3309, as well as some values that are closer to 1, like 1.0042 and 1.0056.

From my very limited knowledge of correlation, seems impossible as I thought these values could only range from -1 to 1. I have been working with a statistician on processing this data, who said the values that are close to 1 (e.g. the 1.0042) are likely due to the complex math involved in getting that number and is likely acceptable. However, I am curious if this is actually true, and further, what it means that I have values that are well over 1 (e.g. 1.4493).

Specifically, I am wondering if 1) these values are acceptable and that I am just interpreting the "meaning" of these values incorrectly, or 2) the values >1 are due to bad data and/or errors in preprocessing that were input to the matrix calculation and should ultimately be excluded from the analysis. Any help or insight is very much appreciated! I have searched this forum and have not located any threads that quite address this question, but it is entirely possible that I might have overlooked something.

Thank you!

Melissa

I was hoping to get some clarification on the outputs from CONN's first-level Seed based analysis (SBC). I exported the conditionxxx.mat files for each of my subjects/conditions and pulled the Fischer transformed correlation coefficients (the "Z") in order to run my own statistics in another program. I've noticed that I some few values that are >1; for example, I have values as high as 1.4493 and 1.3309, as well as some values that are closer to 1, like 1.0042 and 1.0056.

From my very limited knowledge of correlation, seems impossible as I thought these values could only range from -1 to 1. I have been working with a statistician on processing this data, who said the values that are close to 1 (e.g. the 1.0042) are likely due to the complex math involved in getting that number and is likely acceptable. However, I am curious if this is actually true, and further, what it means that I have values that are well over 1 (e.g. 1.4493).

Specifically, I am wondering if 1) these values are acceptable and that I am just interpreting the "meaning" of these values incorrectly, or 2) the values >1 are due to bad data and/or errors in preprocessing that were input to the matrix calculation and should ultimately be excluded from the analysis. Any help or insight is very much appreciated! I have searched this forum and have not located any threads that quite address this question, but it is entirely possible that I might have overlooked something.

Thank you!

Melissa

## Threaded View

Title | Author | Date |
---|---|---|

Melissa DiFabio |
Feb 23, 2021 | |

Alfonso Nieto-Castanon |
Feb 26, 2021 | |