/sci/ - Science & Math » Thread #12354102

194KiB, 922x715, 1601106806070.jpg

View Same Google iqdb SauceNAO

Anonymous Tue 17 Nov 12:38:34 2020 No.12354102 View Reply Original Report

Quoted By: >>12354123 >>12354822 >>12354936

What would be the best statistical models for measuring the relationship between gender and academic performance ?

Anonymous

View Same Google iqdb SauceNAO 1591037391712.jpg, 115KiB, 473x473

Anonymous Tue 17 Nov 2020 12:47:18 No.12354123 Report

Quoted By: >>12354137 >>12354362 >>12354677

>>12354102
Logistic regression or a Neural network maybe?
If you'd be looking at performance then predicting gender then a binary classification Neural Network would probably be best.
Else just use Nadam and a Logistic Regression algorithm unless your dataset is massive

Anonymous

Anonymous Tue 17 Nov 2020 12:49:05 No.12354128 Report

Quoted By: >>12354137 >>12354916

Linear regression, with gender on the X axis and academic performance on the Y.

Anonymous

Anonymous Tue 17 Nov 2020 12:53:48 No.12354137 Report

Quoted By:

>>12354123
Dataset not really massive, live on a small island kek. We're looking at male underperformance. Thank you for your answer.

>>12354128
You too :^)

Anonymous

View Same Google iqdb SauceNAO 1603529534910.png, 139KiB, 649x539

Anonymous Tue 17 Nov 2020 14:21:06 No.12354362 Report

Quoted By: >>12354628

>>12354123
Just use a neural network!

Anonymous

Anonymous Tue 17 Nov 2020 15:43:31 No.12354628 Report

Quoted By: >>12354677

>>12354362
Literally takes an hour at most to set up in a case like OPs unless you're grossly incompetent.
Even then you can just use Googles Tensorflow if its too much of a hassle.

Anonymous

Anonymous Tue 17 Nov 2020 16:02:16 No.12354677 Report

Quoted By:

>>12354628
>>12354123
Neural networks are literally useless for proving statistical correlations/relationships. How about you learn actual statistic instead of relying on computer magic for everything? Fuckwit.

Anonymous

Anonymous Tue 17 Nov 2020 16:37:45 No.12354822 Report

Quoted By:

>>12354102
Logit is good enough. Remember the GIGO principle: no fancy model is going to make up for bad data and bad analysis. Got any covariates or instruments you can use?

Anonymous

Anonymous Tue 17 Nov 2020 17:02:42 No.12354916 Report

Quoted By:

>>12354128
>Linear regression
I like this... keep it simple and clear. If you want more complexity OP go for a PCA.

Anonymous

Anonymous Tue 17 Nov 2020 17:07:59 No.12354936 Report

Quoted By: >>12354944

>>12354102
logitic regression to see if academic performace (grades) predicts gender (a binary variable)

Anonymous

Anonymous Tue 17 Nov 2020 17:09:15 No.12354944 Report

Quoted By:

>>12354936
*logistic sorry I can't type

Anonymous

Anonymous Tue 17 Nov 2020 17:41:59 No.12355122 Report

Quoted By: >>12355139

Thanks for the replies everyone. Logistic regression seems to be the prime choice but I don't know if my colleagues would want gender to be the response variable (whereas with gender as the independent we'd use ANOVA)

Anonymous

Anonymous Tue 17 Nov 2020 17:46:03 No.12355139 Report

Quoted By:

>>12355122
You might want to read up on the theory before doing any regression. Anyway, a gender dummy will be your response variable and the academic performance measure will be your independent variable in a logit. You might want to control for other factors if you want the regression to say anything useful.

Capcode	All Only User Posts Only Moderator Posts Only Admin Posts Only Developer Posts
Show Posts	All Only With Images Only Without Images
Deleted Posts	All Only Deleted Posts Only Non-Deleted Posts
Ghost Posts	All Only Ghost Posts Only Non-Ghost Posts
Post Type	All Only Sticky Threads Only Opening Posts Only Reply Posts
Results	All Grouped By Threads
Order	Latest Posts First Oldest Posts First

Your latest searches