Agree.

From a government spending perspective, I think most first-world countries (and Russia) are good.

Where it seems I differ is in assuming government advantage equates to my personal experience/advantage.

I don't think politicians care about me or my family

Agree.

From a government spending perspective, I think most first-world countries (and Russia) are good.

Where it seems I differ is in assuming government advantage equates to my personal experience/advantage.

I don't think politicians care about me or my family

Why do third world countries have to worry?

Worry about what?

It’s also a questionaire. I took a dump the other day and it was runny. I had Covid in August. Do I have long covid?

Survey says yes

Christ you truly are retarded.....

Create symptoms so vague that they could be a WebMD piece and you get an easy covid bingo for more shots, tests, and everything else that has accomplished... nothing.

Except make covid groupies.

He's either a troll or the long covid has turned his brain into marble.

He’s 100% confirmed troll.



Just remember, this is how he acts even in a thread where he agrees with every poster that a study is bullsh!t and the findings of said study are false.


https://forum.obnoxiousbrutes.com/showt...hp?t=184307613



Imagine ruining over three years of hard work just because you overplay your hand in a thread about a study that even the biggest retard could see is terrible.

The study used multi-variate analysis and proportional hazard models in deriving an association between red meat and diabeetus.



Which would mean they've adjusted for obesity and lifestyle. Scientists are actually aware FYI that if you don't account for random noise your studies are kind of worthless, and it wouldn't pass peer reviews.

Of course, I doubt that will prevent any of the cultists chanting " ZEE JOOZ!" or "MUH PHARMA" or some other sort of ludicrous conspiracy that some evil entity is in control of "science", but hopefully it makes one person think an d consider the study.

But they didn't. You can see in the data tables they did a pisspoor job of adjusting for lifestyle and other dietary factors. It's surprisingly bad.

They conditioned on those variables and reported the effects from those adjusted models.

Without knowing how they “supervised” their machine learning model and how they programmed the analysis tools it’s impossible to know what exactly they are determining in that study, regardless of how many fancy words they throw around.

“ We examined the predictive capacity of derived imaging markers employing a supervised machine learning approach (Fig. 5).”



We also don’t know what their python code looks like:

“All statistical analyses were conducted in Python 3.9.1”

However, when seeing what types of competing interests there are it can provide insight into what they may have “steered” their program towards concluding

That's how MRI analysis works and the tools are cited.

Supervised machine learning uses labeled (ground truth) data to train the model.

If you want to review their code, it's here:
https://github.com/csi-hamburg/2022_...covid_imaging/

Were these the people “supervising” the machine learning to train the model??

No.

They provide all of the code and data.

Who supervised the training of the machine learning?

Regression is supervised machine learning.

Who created and/or selected the algorithm?

I don't think a script will do it.

But I also don't think the techniques used in this study were exotic or novel. Pretty standard rly.

That was only part of the study anyway, evaluating predictive ability that held up with sensitivity analyses.

from sklearn.linear_model import LogisticRegressionCV
from sklearn.model_selection import cross_validate, cross_val_predict, StratifiedKFold
from sklearn.preprocessing import StandardScaler
from sklearn.pipeline import Pipeline
import pingouin as pg

Nothing exotic, custom, or fancy - linear and logistic regression functions from a standard library.

And yet you still haven’t answered the question “who created the algorithms?”

History
This project was started in 2007 as a Google Summer of Code project by David Cournapeau. Later that year, Matthieu Brucher started work on this project as part of his thesis.

In 2010 Fabian Pedregosa, Gael Varoquaux, Alexandre Gramfort and Vincent Michel of INRIA took leadership of the project and made the first public release, February the 1st 2010. Since then, several releases have appeared following a ~ 3-month cycle, and a thriving international community has been leading the development.

Scikit-Learn is a community driven project, however institutional and private grants help to assure its sustainability.

The contributors, developers, and funders are listed on the same page as this info

https://scikit-learn.org/stable/about.html

Source code is available here

https://github.com/scikit-learn/scikit-learn

Not really surprised, such is the life of a beavercel. And I agree….I’d be extremely interested to hear about any conflicting interests he has.




And since I won’t be discussing the code with him due to his insistence on lying about what people say, I guess I’ll just share this with you and others who may be interested.


https://github.com/csi-hamburg/2022_...ediction.ipynb



There are a few potential ways in which this code could output inaccurate data:

1.) Inconsistent or incorrect data preprocessing: The accuracy of a machine learning model is highly dependent on the quality of the input data. If the prediction_df used to create the X and y variables is not properly preprocessed, it may contain missing values, outliers, or other inconsistencies that could lead to inaccurate results.

2.) Insufficient cross-validation: The code uses nested cross-validation to evaluate the model's performance. However, the number of iterations (iterations=100) may not be sufficient to obtain stable and reliable results. Increasing the number of iterations could help improve the accuracy of the output.

3.) Improper parameter search: The code uses logistic regression with elastic net regularization and performs parameter search using LogisticRegressionCV. The choice of parameter values and the range of values to search may not be optimal for the given dataset. Adjusting the parameter search space or using alternative parameter optimization techniques could potentially lead to more accurate results.

4.) Randomness in cross-validation: The StratifiedKFold cross-validation strategy is used with the shuffle=True option and a different random_state value for each iteration. The random splitting of data could introduce variability in the results. To obtain more reliable estimates, you may consider fixing the random_state to a specific value or performing multiple runs with different random seeds and averaging the results.

5.) Overfitting or underfitting: The code does not include any measures to prevent overfitting or underfitting of the model. It is possible that the model is either too complex or too simple for the given data, leading to inaccurate predictions. Regularization techniques, such as adjusting the regularization strength or exploring other models, could help address this issue.

But the data and code are all freely available.

They used an an accurate, but sophisticated sounding, term for their predictive regression. Probably not necessary, but certainly not a gotcha or reason to disregard the results.

If we consider any study using linear regression (by extension ANOVA) as invalid, we don't have much left. Regression is kinda what everyone does.


The code is well organized and runs quick. You can see for yourself if any of these concerns are present in this data set.

Suggesting there is no code to address or assess under/over fitting while also noting their CV is a little sus.

Just fukkin lmao at how upset Justin’s Little Cuck gets when exposed as a troll.