WM volumes of the SCG will be the continues variable. Due to potential differences between the WM volumes of men and women, the effect of gender will be assessed.



Key question:

-Are there differences in volume between men and women.

- And if there are differences. Is this explained by age?



WM volumes of the SCG will be the continues variable. Due to potential differences between the WM volumes of men and women, the effect of gender will be assessed. 

When the 93 manual parcellations for the subjects are done it is of importance to see if data are normally distributed and if there is homogeneity of variances. This will be analysed by a histogram and a residual plot. The results of the Kolmogorov Smirnov and the Shapiro-Wilk will be used to assess normality. If these assumptions hold, an independent samples t-test will be performed. If not, a non-parametric test such as the Mann-Whitney test will be executed or the data will be transformed, when appropriate. For example, with a log transformation depending on the skewness of the data.

In addition, it is of importance to asses if there are significant differences in the volumes of the SCG between the LH and RH. For this a paired t-test will be used. If there is no difference between the SCG of the LH and RH the data will be collapsed to form data of one individual.

Based on the results of the above described analyses, it will be determined how the dataset will be treated. If there are significant differences between de LH and RH, the datasets for each hemisphere will be analysed separately of each other. If the datasets of the LH and RH do not differ, the datasets can be treated as one dataset.

Furthermore, because of the possibility of differences in age between male and female this will also be assessed with an independents t-test. Afterwards, if the necessary assumptions hold, a regression model will be constructed containing age and gender as explanatory variables or a stratified analysis on age groups will be performed.


Please do the above described for the 93 subjects done by myself.

In the single datafile you can recognize the 93 subjects done by myself by my initials YTT in the name of the datafile: example sub-001_ses-1_acq-wb_mod-t1map_mask-scg_hem-l_rat-ytt_calc-bin_out.nii.gz


Then repeat the same but with adding the remaining 14 subjects.

The remaining 14 subjects were excluded in the previous analysis for my part of parcellations but can be included for a second analysis by adding the 14 single others, recognizable by the initials of my supervisor AAX:

example sub-002_ses-1_acq-wb_mod-t1map_mask-scg_hem-l_rat-aax_calc-bin_out.nii.gz








PLEASE! Check the usual assumptions of the data first and also write something about this (so the distribution of the data, sphericity, outliers and all the usual stuff)

I already had did some analyses concerning the outliers but lost the files unfortunately and I can already report that if I wrote it down correctly, in the single datafile for the Left Hemisphere the volumes for scanid-066 and scanid-095 are outliers and should be removed cause they are human errors made by me (did not save these files correctly). For the right hemisphere sub-038 information is missing so this right hempishere should also be excluded. Furthermore, for the right hemisphere the data for the following participants, scanid-065, scanid-086 & scanid-014 should also be outliers but can be included in the analysis because they are not human errors, their subcallosal gyrus just happens to be big. But please do the analyses again to see if you get the same outliers or different ones? Let me know ASAP if there are any other outliers so I check the brain volumes files immediately in another software named FSLeyes to see if this are human errors and should be excluded.


Very important! These are the analyses I want to have done. Please with tables.

1.      Are there gender difference in the volumes of the SCG in the 93 subjects done by me YTT (in the filename). So, every scanid in the single file except the following 002, 007, 012, 017, 028, 031, 035, 043, 044, 047, 055, 059, 065, 092 if I’m correct.

2.      Are there gender differences in the volumes of the SCG in the 107 subjects done by me (YTT) and AAX).


The above analyses are quite simple and you can find the data in the file named ‘single’. The following analyses are conjunctions. You can find them in the datafiles named ‘conjunctions’.  Please check again the same same assumptions again described in the black colored text above.


3.      Is there a gender difference if we make an mean for every hemisphere for every subject (Segvolume+Refvolume = outcome, outcome divided by 2 for every participant/hemisphere). So, we can say we made the sample more trustworthy by adding multiple raters in to the analysis. Afterwards please analyse if there is an effect for gender.


I will try to explain what this file consists of but I don’t know the exact details to be honest cause my supervisor made this rapport for me:

1.      Segvolume: This is the ‘primary’ volume. It doesn’t really mather but if you want to know how this variable existed. The parameter for Segvolume is very simple. If rater AAX did the segmentation include this segmentation in Segvolume if AAX did not do the segmentation, use the segmentation of YTT. The thought behind this is that AAX (my supervisor) has more experience in parcellations of the SCG then me, therefore here parcellation is probably of higher quality and if possible, should always be included if hers is not available I am so to speak the second preferable rater cause it’s my research.

2.      Refvolume: After the segmented volumes were determined we need another segmentation of the SCG to make conjunctions. For this variable the parameter is also simple. The code first checked if there is a segmentation available from me as a rater (YTT) if this was not available a third rater (also a student as me) came in to play with the initials VHX. So when Segvolume was done by AAX the chosen Refvolume was done by YTT. But if YTT already was chosen for the afore mentioned Segvolume different rater should be choosed and this was a third rater (VHX) his segmentations were used as Refvolume to make conjuctions with. This was done to calculate the following variables.

3.      Volume difference. I really don’t understand this variable and don’t know which analyse she used herefore. For example: Sub-002 has the following Segvolumes and Refvolumes.






I don’t know which analyse type she used here to come to the following volume difference. Maybe for now we should ignore this part unless you seem to know which analyse type she used? Also I don’t know how this should give any information about my key questions?



4.      Dice overlap. This one is very important! This analyse tells how good inter-rater agreement is. How higher the overlap how better the segmentations of AAX VS. YTT or YTT VS. VHX how better the inter-rater agreement. If you want more info about the dice score check:


It’s important that you analyse the Mean and SD of this dice score.


5.      Average surface distance

What I found about this: https://lh5.googleusercontent.com/ZYFVFeutahduTobPrDacOLlYTEfsVZyfQ1OsGp87P-Sx9O8-TiBxs8GeNHXpjNmaW-DO3lykEPfNgLor5FfcXhDnaOJK8CUC0wA0LAwpPE9YlU5IVef16Pt_upolwdBD8mpzoLyh

What I understand of this is that the average surface distance is the Mean Error between de distance of the contours of the mask (segmentation of the SCG) and the voxels. So how lower the value of this error how better the inter-rater agreement. Please also do an analyse on the mean and sd of this and find an useful article of this and cite this while telling something about the mean of this error.


6.Possible outcomes could also be explained by another factor, namely age. Please also check this for me.


7. Last but not least! Please write everything down that you do. So which tests you did and which participants you excluded or not.


Related Questions in statistics category