Karolis Koncevičius
  • Member for 9 years
  • Last seen this week
How to plot binary (presence/absence - 1/0) data against continuous variables
Accepted answer
25 votes

If I understood the question correctly - you might want to use a "conditional density plot". Such a plot provides a smoothed overview of how a categorical variable changes across various levels of ...

View answer
When should I apply feature scaling for my data
21 votes

In my view the question about scaling/not scaling the features in machine learning is a statement about the measurement units of your features. And it is related to the prior knowledge you have about ...

View answer
What happens in the sub-areas of AI? (ML, DL)
16 votes

This is just terminology, no need to think about it too much as different people classify different areas into different categories. For example a lot of statisticians would consider machine learning ...

View answer
What does standard deviation mean in this case?
12 votes

Interpretation of the Mean When we say that the average value spent on meals was 50 USD - it means that if we take the total amount spent on fast foods and equally divide the sum among all the people ...

View answer
Name of classification algorithm based on gaussian distributions estimated from data?
Accepted answer
12 votes

Probably Quadratic Discriminant Analysis. There are also names for different constraints you could make: Covariance matrices of both classes are equal - Linear Discriminant Analysis. Only diagonal ...

View answer
Should I report the descriptive statistics in publication before or after outliers removal?
12 votes

In publications honesty is the best policy. It's all right to remove outliers and to report descriptive statistics based on the reduced dataset. However in order for others to understand and ...

View answer
Cross validation confusion on the concept
Accepted answer
11 votes

In simplified terms you can think about it like this: suppose you are preparing pupils for an exam. You have three sets exercises from previous exams: A, B, and C. But the exercises in the upcoming ...

View answer
Recommended terminology for "statistically significant"
9 votes

In my opinion one of more honest yet non-technical phrasing would be something like: The obtained result is surprising/unexpected (p = 0.03) under the assumption of no mean difference between the ...

View answer
Normalize an array of numbers to specific range
8 votes

@NickCox already provided an answer in the comments, I will only elaborate on it step-by-step and try to provide some intuition. You have a range on numbers ranging from $x$ to $y$ and you want to ...

View answer
I've already used my entire dataset in a regression, should I not use that as a prediction model?
7 votes

In my opinion the best course of action, if possible, is to collect more data and then use that data to check your current model as well as maybe top 5 of the previous models you tried. Continuing ...

View answer
Usefulness of the confidence interval
7 votes

I think it's best understood with a simple example. Imagine you are on a farm that raises sheep. The farm has a lot of sheep but you only observe 5. Out of those 5 sheep, 1 sheep was black and 4 sheep ...

View answer
How to identify if a problem is a good candidate for applying machine learning?
7 votes

Prof. Yaser Abu-Mostafa talks briefly about this in his Caltech course on machine learning during the first lecture. He identifies 3 essential points you have to consider before considering applying ...

View answer
Why linear discriminant analysis is sensitive to cross validation (LDA overfit problem)?
7 votes

Looks like your sample size is not a lot bigger than the dimensionality of the data (feature set size). That can be a problem for LDA and it can overfit. Since it relies on computing the within-class ...

View answer
Does SVM suffer from curse of high dimensionality? If no, Why?
6 votes

SVM also suffers the problems coming from high dimensionality, but under typical settings to a lesser degree compared to (say) LDA. I can imagine SVM would only have to take dot products of support ...

View answer
Best practices in the selection of distance metric and clustering methods for gene expression data
Accepted answer
5 votes

This will probably not be the answer you want or expect, but this is how I see these things. Clustering problem Clustering, to a degree, is almost always a subjective procedure. You decide how you ...

View answer
What test exactly does lm.anova perform in R?
Accepted answer
5 votes

R has comprehensive documentation. For this specific case help(anova.lm) says: Details: Specifying a single object gives a sequential analysis of variance table for that fit. That is, ...

View answer
Preprocessing gene expression dataset
5 votes

Strictly speaking - there is no proper way to do this. It depends on your problem and more specifically - what you know about your dataset and what you are trying to achieve. If, for example, in your ...

View answer
Why do Experimental Design when you could build classifiers?
5 votes

What is the difference between constructing a model vs hypothesis testing? Like @mkt mentioned - both of these involve models in one for or another so that's a false distinction between them. ...

View answer
Find the rotation between set of points
5 votes

This can be done using the Kabsch Algorithm. The algorithm finds the best least-squares estimate for rotation of $RX-Y$ where $R$ is rotation matrix, $X$ and $Y$ are your target and source matrices ...

View answer
Three questions about the article "Ditch p-values. Use Bootstrap confidence intervals instead"
4 votes

1. Citizenship example This seems to be a poor but valid test. It makes sense with an extreme p-value cut off $\alpha = 0$. This way we would only reject those citizens whose profession is not found ...

View answer
Why does error rate of kNN increase when k approaches size of training set?
Accepted answer
4 votes

Adding my guess as an answer. When splitting the data into train/test sets you do a 0.8/0.2 split. Which, with the sample size of 500, should be 400/100. Then you train kNN on 400 and test the ...

View answer
Why does machine learning have metrics such as accuracy, precision or recall to prove best models, but statistics uses hypothesis tests?
4 votes

Anything can be seen as a "metric", and both groups, statisticians and machine-learners, use plenty of those: accuracy, mean value, estimated parameter of a model, etc. Hypothesis testing is ...

View answer
How to report t statistic from R
4 votes

If you have to report all the details then you should also report the actual t-value, not just degrees of freedom. About the degrees of freedom: your degrees of freedom changes because you are using ...

View answer
What is exact 'learning' on a deep learning model?
Accepted answer
4 votes

The question is a bit unclear, but here is an attempt at an answer based on my understanding of the question: First thing that might hinder some understanding is the (in my opinion) false distinction ...

View answer
Why do we need alternative hypothesis?
4 votes

Im am not 100% sure if this is a formal requirement but typically the null hypothesis and alternative hypothesis are: 1) complementary and 2) exhaustive. That is: 1) they cannot be both true at the ...

View answer
Avoiding social discrimination in model building
4 votes

This at most will be a partial answer (or no answer at all). First thing to note is that I agree with @dsaxton completely: all models "discriminate" (at least in some definitions of discrimination) ...

View answer
Why do we plot the standard error of the mean as both positive and negative?
Accepted answer
4 votes

Because your estimated mean might differ from the true mean in either direction: it can be too high or it can be too low. If you estimate the mean from your sample to be (e.g.) 3 and you also know ...

View answer
3 D6 pick 2 vs 2 D6 with 1 reroll
Accepted answer
4 votes

If the goal is to obtain 7+ eyes and there is no preference about the total score otherwise - then there is no difference between rolling 3 vs rolling 2 and re-rolling the lower one when their sum is ...

View answer
Why is it so important to have principled and mathematical theories for Machine Learning?
4 votes

Empiricism vs Theory You wrote: One of the biggest criticism of theory is that because its so hard to do, they usually end up studying some very restricted case or the assumptions that have to be ...

View answer
power analysis of a proposed study
4 votes

The short answer is that you can't. Like other answers mentioned you need to assume a fixed effect size, fixed significance level and a desired statistical power in order to estimate the needed sample ...

View answer