zkurtz
  • Member for 8 years, 6 months
  • Last seen more than a week ago
  • Pittsburgh
How is it possible that Poisson GLM accepts non-integer numbers?
20 votes

Of course you are correct that the Poisson distribution technically is defined only for integers. However, statistical modeling is the art of good approximations ("all models are wrong"), and there ...

View answer
Can I compare the p-values of two Wilcoxon tests?
Accepted answer
7 votes

No. According to Gelman and Stern (2012), "even large changes in significance levels can correspond to small, nonsignificant changes in the underlying quantities." In plain English, differences ...

View answer
Understanding the Rank Probability Score
Accepted answer
6 votes

Equation 7 in Hersbach inspires me to notice that the RPS (as a discrete version of the continuous RPS, or CRPS) is a sum of several Brier quadratic probability scores (BS) evaluated over several ...

View answer
Cohen's Kappa using (irr) and kappa2() outputs NaN
5 votes

In this data, your two raters agreed 100% of the time (the function excludes all rows that include an NA). They both individually have a 100% estimated chance of selecting category "1". Thus, the ...

View answer
What's the best Kernel Regression package in R?
5 votes

Of course which one is the "best" depends on some rather subjective things, but the gam command in the mgcv package meets all of the criteria you described.

View answer
t-test on highly skewed data
4 votes

When n is large (like 300, even far less than 3000), the t-test is essentially the same as the z-test. That is, the t-test becomes nothing more than an application of the central limit theorem, which ...

View answer
Prediction Intervals with Heteroscedasticity
4 votes

Nonparametric quantile regression gives a very general approach that allows for both heteroscedasticity and nonlinearity. See section 9: http://www.econ.uiuc.edu/~roger/research/rq/vig.pdf UPDATE: ...

View answer
Can the Burnham-Anderson book on multimodel inference be recommended?
4 votes

The OP appears to be seeking a high-quality survey of high-quality statisticians to help assess whether one particular book is of high quality particularly with regards to the AIC versus AICc debate. ...

View answer
Marginal effect in model with interactions
4 votes

You seem to be aware that the marginal effect of $X_3$ is $\delta_1 + \delta_2 X_1 + \delta_3 X_2$, which is just the derivative of the response with respect to $X_3$. Replacing $X_1$ with $E(X_1)$ ...

View answer
Inconsistent results using Bayesian Model Averaging for model selection?
Accepted answer
4 votes

The short answer is that there is no particular reason to be surprised at what you are seeing. The use of p-values to decide which variables to include in a model is rather outdated, and most ...

View answer
When does a distribution not have a mean or a variance?
Accepted answer
3 votes

The canonical example of a distribution with no mean (and hence no variance) is the Cauchy distribution. True, you could compute the sample mean for a sample of data from a Cauchy distribution, but ...

View answer
Rare event logistic regression bias: how to simulate the underestimated p's with a minimal example?
Accepted answer
2 votes

Figure 7 in the paper seems to most-directly address the question of bias in the predictions. I don't fully understand the figure (specifically, the interpretation "estimated event probabilities are ...

View answer
Structural Equation Models (SEMs) versus Bayesian Networks (BNs)
2 votes

I don't really understand this, but see here: Structural equation models and Bayesian networks appear so intimately connected that it could be easy to forget the differences. The structural ...

View answer
Bootstrap test for correlation coefficient
Accepted answer
2 votes

Just sample pairs (X,Y) with replacement. You should not be permuting anything.

View answer
Normal distribution for discrete values
2 votes

Taking the question at face value, the answer is zero. The area under the standard normal density curve at any single point is the area of a line segment, which is always zero.

View answer
Applying Fisher's test to my data
2 votes

Rate A is 4/20 = 0.2. Rate B is 6/10 = 0.6. By "Is this statistically significant" I presume you mean, "is the difference between rate A and rate B statistically significant?" (And not "is the ...

View answer
Does Cohen's Kappa sample size calculation need a check for "true kappa" feasibility?
1 votes

Yes, it should fail with a warning. The reasons cited in the OP are correct; here I'll simply derive an upper bound for kappa that proves the point. Define several "true" population level parameters:...

View answer
Specifying lag in `dlnm` when passing arguments to `crossbasis`
1 votes

This answer will likely be an iterative process, because you may want to rephrase your question as things become more clear. 1) In your usage, the first argument to crossbasis is a vector. According ...

View answer
Determining sample size for a comparative study involving two unequal groups
1 votes

The answer is 11. Given that the size of the first group is 10 times the size of the second, 11 is the smallest possible sample size that includes at least one observation. Now, you might say that ...

View answer
Is it possible to compare time series method with non-time series method?
1 votes

Yes. A non-time series method that I have used (as a "hack") to model time series is a generalized additive model (GAM) with a nonparametrically-smoothed time effect. For example, using the mgcv ...

View answer
Measuring relationships between variables directly (survey)
Accepted answer
1 votes

You have identified most of the important issues. As you stated, direct measures of "CrossValidated usage" or "team productivity" may be difficult. Perhaps an even deeper issue is this: even if you ...

View answer
Using AIC to select between models that use nested and non-nested variables
Accepted answer
1 votes

According to an informal document by Burnham, who I regard as one of the leading experts on the AIC, the notion that models need to be nested to use the AIC for model comparison is a myth. Here is ...

View answer
When can we look at the significance of pairwise comparisons rather than an interaction term?
Accepted answer
1 votes

To paraphrase part of your approach, your are doing two pairwise comparisons, resulting in two p-values, say p1 and p2. If p1 is small and p2 is not small, you would like to conclude that the ...

View answer
Data mining for a statistician
0 votes

The short answer is that you seem to be on the right track. If you are on the wrong track in some way, it is because you seem to be searching for a precise definition for data mining, when in fact ...

View answer
Regression of data that includes a date
-1 votes

Pick a reference date, say 1/1/2010, and make a new variable time that is the difference between the date and the reference date, where the difference is computed in, say, days. Now run a linear ...

View answer
Absolute values -- parametric or non-parametric test?
-1 votes

The first thing to understand is that the assumption of normality when using the t-test can never be proven. One can only say that the data look approximately normal, and if this is the case, the t-...

View answer