Reputation
918
Top tag
Next privilege 1,000 Rep.
See votes, expandable usercard
Badges
5 13
Newest
 Tumbleweed
Impact
~65k people reached

Aug
16
comment Is there a multiple testing problem when performing t-tests for multiple coeffcients in linear regression?
Sorry I don't have time right now to give you a better answer, but the results from the regression are answering a particular hypothesis test (comparing distributions under the inclusion or removal of a particular variable. A Bonferroni-like adjustment would instead be more appropriate if the null hypothesis was more complicated including a vector of parameter. This is the case in multiple comparisons.
Jun
9
comment Fitting the differences between two curves
Let us continue this discussion in chat.
Jun
9
comment Fitting the differences between two curves
I'm still not happy with the definition; maybe a better statement would be along the lines of testing that the difference between two 'lines' are decreasing over a fixed interval. So what is generating these 'lines'?
Jun
9
comment Fitting the differences between two curves
I think you really need to flesh out what convergence means, in explicit language. If two things cross and remain close they can be parallel for a sufficiently small difference between each other, you could also have to coincident vertical lines, they clearly converge to each other, but diverge in several definitions of convergence. Is anything here random?
Jun
9
comment Fitting the differences between two curves
Can you define converge and diverge in a slightly less visual sense, and what the properties of these curves are? Remember convergence is defined as one object becoming another object in some sense such as the convergence of a sequence of values to another value, or convergence in distribution of a random variable to another random variable.
Jun
3
comment Likelihood-based hypothesis testing
You might want to clarify this a bit, you first state that you know the parameters, and you don't have an estimate. Since you know the parameters you already know if they are the same or not, so why are you doing a test? Then you start talking about CI's for parameters, unless you are doing something Bayesian, your parameters are not stochastic. Remember that you do tests based on the distribution of your estimator, and these tests should reflect the chance that the estimator comes from some null distribution.
Jun
3
comment Using SVD or PCA for reducing dimensionality
I would check out the links specified by @amoeba. What you have done here is reduce the column space or rank of $A$, but you haven't changed its dimension. What you want to do is to project your $d$-dimensional space into a $b$-dimensional space $d > b$. This can be done with eigenvectors/PCA, but you might loose some interpretability in your analysis.
May
31
answered How a statistical package like SAS analyses market risk without any calculus support
May
31
comment Best classifier for data with text among features
You might want to avoid using the word 'best', as it implies optimal which is unlikely to be achieved in your case (this word has very specific meaning in statistics, usually optimal under a squared error loss function). Furthermore you might want to give an example of a few observations including features and the label; in particular I find "label 2 of the features is a text" what does this mean, and what is 'text"?
May
22
comment Feature extraction based on correlations
How many voxel's are you looking at this could quickly get computationally intractable, quite quickly. Efficiency in this context is also a function of how much time are you willing to wait, and what computational resources you have access too. Intuitively I don't see what you are really trying to do with these correlations, what does the PCA of these things do for you?
May
17
answered How to reset options to default in SAS 9.3?
May
17
revised Calculate Significance between samples with skewed data
html broken
May
17
comment Calculate Significance between samples with skewed data
I'm just a bit confused about why you are trying to subsample, since you already have a larger sample. Are you trying to determine sample size for a future sample?
May
17
revised reasons for a non-invertible matrix
added an example
May
17
answered reasons for a non-invertible matrix
May
16
awarded  Tumbleweed
May
16
comment Rescaling Features for ML
The problem with k-means is that no matter how you scale the data, you will always have issues with outliers. Using a log transform should work, but this may change the relationships between your variables considerably. You may also want to take a look at k-medoids or other more robust classifiers.
May
14
comment How to make sure that a machine learning algorithm's implementation is correct?
That was sort of the point. If you have a black box with a few examples and no binary to deconstruct, you are frankly out-of-luck. If this is a 'novel' method and there is a paper you can reproduce the program to the best of your ability, and that's about it. If you have a binary, but can't reproduce the results because of some initial values, you should at least know what object the initial values are (vector, scalar), and check how sensitive the algorithm is to choice of initial values.
May
14
answered Calculate Significance between samples with skewed data
May
14
comment What are the real benefits of normalization (scaling values between 0 and 1) in statistics?
A lot of the answers are with respect to normalization (standardization), if this isn't what you are looking for could you please expand on how you are normalizing (scaling) your data. Are you just scaling the score range over the interval 0,1?