User Josh

10 votes

2 answers

4k views

How to avoid overfitting bias when both hyperparameter tuning and model selecting?

asked Nov 3, 2020 at 14:24

9 votes

2 answers

2k views

What are the benefits of time-series over a well-setup linear regression for forecasting?

asked Jun 25, 2021 at 21:20

3 votes

0 answers

541 views

Area under the lift curve as a model performance metric?

asked Dec 11, 2021 at 18:24

3 votes

1 answer

119 views

How to efficiently estimate number of individuals with n+ successes from a series of bernoulli trials?

asked Apr 24, 2022 at 0:07

3 votes

1 answer

536 views

Does gradient boosted trees actually use regression trees for classification, and if so, what does the gradient update?

asked May 28, 2021 at 17:15

3 votes

2 answers

352 views

What is accepted practice for avoiding optimistic bias when selecting a model family after hyperparameter tuning?

asked Nov 16, 2020 at 22:01

3 votes

1 answer

61 views

How to balance transformation decisions, feature selection, and model tuning vs time in text analytics?

asked Oct 19, 2020 at 17:31

1 vote

0 answers

166 views

Bootstrap and distribution under the null for confidence intervals vs p-values

asked Sep 24, 2019 at 20:56

1 vote

2 answers

329 views

Bootstrap for CIs and Permutation Resampling for Hypothesis Test?

asked Sep 26, 2019 at 14:27

1 vote

0 answers

46 views

Is there an analogous calculation for weighted max and min, similar to weighted average, for text embeddings?

asked Jan 21, 2023 at 16:02

1 vote

2 answers

648 views

How to calculate a sense of error for estimated number of events with known probabilities?

asked Mar 12, 2022 at 1:07

0 votes

1 answer

52 views

Is it possible to efficiently compute the probability of k+ events in a Poisson Binomial Distribution?

asked Dec 13, 2022 at 17:47

0 votes

0 answers

26 views

Conceptual understanding of effect of standardizing after normalization for clustering

asked Feb 24, 2023 at 19:43

0 votes

1 answer

430 views

Nested CV vs CV with a holdout?

asked Nov 3, 2020 at 13:41

0 votes

0 answers

175 views

Multivariate, multi-series time data more appropriate for regression or ARIMA/time-series?

asked Apr 10, 2021 at 11:57

0 votes

1 answer

137 views

Why is the point biserial correlation of numeric variable and its binary, binned self not 1.0?

asked May 5, 2021 at 23:55

0 votes

0 answers

139 views

Will ranking "correlation" of variables by Pearson's return the same results as chi-square and F stats?

asked May 10, 2021 at 21:40

0 votes

0 answers

34 views

Is balancing class data for imbalanced problems helpful or just folklore when considering thresholds?

asked May 19, 2021 at 12:18

Stack Exchange Network

18 Questions

How to avoid overfitting bias when both hyperparameter tuning and model selecting?

What are the benefits of time-series over a well-setup linear regression for forecasting?

Area under the lift curve as a model performance metric?

How to efficiently estimate number of individuals with n+ successes from a series of bernoulli trials?

Does gradient boosted trees actually use regression trees for classification, and if so, what does the gradient update?

What is accepted practice for avoiding optimistic bias when selecting a model family after hyperparameter tuning?

How to balance transformation decisions, feature selection, and model tuning vs time in text analytics?

Bootstrap and distribution under the null for confidence intervals vs p-values

Bootstrap for CIs and Permutation Resampling for Hypothesis Test?

Is there an analogous calculation for weighted max and min, similar to weighted average, for text embeddings?

How to calculate a sense of error for estimated number of events with known probabilities?

Is it possible to efficiently compute the probability of k+ events in a Poisson Binomial Distribution?

Conceptual understanding of effect of standardizing after normalization for clustering

Nested CV vs CV with a holdout?

Multivariate, multi-series time data more appropriate for regression or ARIMA/time-series?

Why is the point biserial correlation of numeric variable and its binary, binned self not 1.0?

Will ranking "correlation" of variables by Pearson's return the same results as chi-square and F stats?

Is balancing class data for imbalanced problems helpful or just folklore when considering thresholds?