Questions tagged [sequential-analysis]
Hypothesis testing without a fixed sample size.
120
questions
1
vote
0
answers
23
views
Distribution for Fraction of Success in a Binomial Setting
So the actual original question I am trying to solve is a little bit different than the title:
In a binomial setting with probability of success $p$, I keep examining observations until a fraction of ...
0
votes
0
answers
21
views
Expected sample size for sequential t-test
I have a 2 group sequential one-sided t-test setting.
$1-\beta = 90\%$, $\alpha = 5\%$.
Null hypothesis is $\mu_1=\mu_2=1800$.
Alternative hypothesis is: $\mu_1\geq\mu_2$
Plug-in estimate of $\sigma$ ...
2
votes
1
answer
39
views
Detection of 'unfair' Bernoulli sequences using run lengths?
I'm rather confused about the following question:
Which of the following vectors do you think is actually a random sample of $\mathcal{B}(1, 0.5)$ (Bernoulli) of length 10?
$[0,1,1,0,1,0,1,0,1,1]$ and ...
1
vote
0
answers
12
views
Should SPRT use explicit hypotheses or reflect observed differences in the data?
Sequential Probability Ratio Testing (SPRT) compares the log-likelihood ratio of the observed data under the null and alternative hypotheses to thresholds based on desired alpha and beta levels.
I am ...
0
votes
0
answers
22
views
Adjusting effect sizes, p values, and CIs for sequential analysis via simulation
It's fairly easy to write a simulation procedure to calculate, for any given global type 1 error rate, adjusted local "stopping" alphas for interim "looks" in a sequential design, ...
1
vote
0
answers
18
views
How is the O'Brien-Fleming constant calculated and how does the procedure control error rates?
O'Brien-Fleming is a commonly used interim analysis procedure in clinical trials. The procedure is used when $N$ tests are performed in sequence. For each test $n \in \{1, ..., N\}$, a test statistic $...
1
vote
1
answer
25
views
Estimating population means after a selection procedure
The problem is somewhat related to sequential procedures (e.g. Paulson; Gupta and Miescke)
I have a hundred engines generating normal random numbers with true (unknown) means $E_1,E_2,E_3...E_{100}$, ...
0
votes
0
answers
31
views
I do not understand, are Neyman-Pearson rule and Sequential Probability Ratio Test refers to the same thing?
In Bayesian statistics, I remember to Neyman Pearson rule, in which, based on a threshold value, we have the option of reject to make a decision: decision option.
Now, I wonder if it is the same as ...
0
votes
0
answers
12
views
Sequential Probability Ratio Test (Property Proof)
Can anyone provide me the easily explainable proof of 'SPRT terminates with probability unity'?
To be more specific- 'Prove that SPRT terminates with probability unity'
1
vote
1
answer
49
views
How to explain these jump-discontinuities in the expected stopping time of a SPRT?
Suppose a sequential probability ratio test (SPRT) with $$\alpha=P(\text{Type I error})=P(\text{Type II error})=\beta = 10\%$$ is used to decide between $$\ H_0: p=1/2\quad\text{vs.}\quad H_1: p=b\,$$ ...
2
votes
1
answer
34
views
Changing sample size during experiment? [duplicate]
Will the significance and power levels hold constant if I don't see statistically significant results and decide to wait to allow for more samples to be collected?
0
votes
2
answers
59
views
When should a problem be approached as a sequential one?
I am working on a problem with a dataset of the form $(x_j,t_i,y_{i,j} )_{(j,i) \in J \times I}$ where $y_{i,j}$ is the output label. Think of progression of illness over time or weather conditions ...
0
votes
0
answers
51
views
Bayesian sequential updating with current Bayesian sampling software?
I'm having a hard time implementing sequential updating with current software, I don't even know how to start. Basically, I'm having to refit the whole model by simply appending the new data to the ...
3
votes
0
answers
38
views
Sequential hypothesis test for uniform distribution
Suppose I have a sequence of data points $x_1,x_2,\dots$. I want to do a sequential hypothesis test, were the null hypothesis is that each $x_i$ is iid uniform on $[0,1]$, and the alternative ...
1
vote
0
answers
37
views
Sequential Testing of Poisson Process
My question is related to the paper "Sequential Testing for Poisson Processes" by Peskir and Shiryaev, available here.
Specifically, suppose that there are two states of the world, G(ood) ...
0
votes
1
answer
80
views
How can I adjust the p-value calculation for an A/B test that is currently running?
I am running an A/B test. I have about 95k samples per variation and 1.29% relative improvement. The test is to see whether the variant converts users to booking better than the control. Test is still ...
0
votes
1
answer
110
views
Group Sequential Clinical Trial Design - Non-binding efficacy boundaries?
I have been reading up on group sequential clinical trial designs and came across the concept of non-binding futility boundaries. Intuitively, it makes sense to me that continuing the trial after ...
2
votes
1
answer
136
views
Sequential updating of a Bayesian regression model
I'd like to perform Bayesian regression of $Y$ on $X$. i.e. estimate a posterior $p\left(\theta\mid\mathcal{D}\right)$ of some regression parameter $\theta$ given a dataset $\mathcal{D}$ of pairs of ...
8
votes
1
answer
340
views
How to distinguish between two biased coins
I have two very biased coins:
The first coin (C1) lands on heads 95% of the time.
The second coin (C2) lands on heads only 1% of the time.
One of the coins has been selected and I want to determine ...
3
votes
1
answer
156
views
Term for when hypothesis testing is conducted repeatedly until an interesting result is found?
What is the term for when, through accident or intent, many hypothesis tests are conducted until an 'interesting' result is found? That is, the experiment (or variations on it) are repeated until a ...
1
vote
0
answers
47
views
In R, why do the p-values from anova() change when you add more predictors?
When conducting linear regression in R, I am trying to understand how certain p-values are calculated and what they represent. So far this is my understanding:
The p-values from summary() correspond ...
5
votes
1
answer
138
views
Equivalence of Contextual Bandit formulations
I find two different type of Contextual Bandit problem formulations in the literature:
Definition 1: (https://hunch.net/~jl/projects/interactive/sidebandits/bandit.pdf) In a contextual bandits problem,...
0
votes
1
answer
157
views
Is sequential probability ratio test bayesian or frequentist or both?
Is sequential probability ratio test bayesian or frequentist or both?
Is there a way to implement a bayes decision rule with "sprt"?
2
votes
1
answer
51
views
How long will it take to detect that the parameters in a normal distribution have changed, given online sampling?
Let's say, for instance, you have normally distributed data that is coming in one at a time. At some unknown instance, the distribution parameters (mean and variance) change. Assuming that there is ...
1
vote
0
answers
19
views
Early stopping in wilcoxon signed rank test?
I have decided to acquire N joint samples of variables X and Y (for example, N=200), and test whether their means are different using the signed-rank test. I will repeat this procedure many times, ...
1
vote
0
answers
557
views
O'Brien Fleming Interim Analysis Interpretation with PROC SEQDESIGN
Quick question about the output of PROC SEQDESIGN. I was playing around with the O'Brien Fleming method trying to understand how we can interpret the results spit out by SAS for a mock study with ...
2
votes
1
answer
68
views
How to recursively filter sequential predictions of binary classifier?
Problem
My goal is to estimate binary state using entire observation history. Observations are coming sequentially with fixed interval (~1 sec). I have xgboost model which predicts the state given a ...
2
votes
1
answer
91
views
To improve the posterior belief receiving a time-series from a fixed data source
Let data $\mathbf{X}\in \mathbb{R}^d$ come from one of the $K$ possible sources $\mathsf{S}\in \{1,2,...,K\}$. The true $\mathsf{S}$ is unknown but it is fixed. The main task is to infer the true ...
1
vote
1
answer
33
views
extension of sequential probability ratio tests to particle filters?
I've been wondering if there are extensions of the sequential probability ratio test to account for particle filters. I ask because, in my research, I'm working with distributions that cannot be ...
0
votes
1
answer
99
views
Binary Sequence Prediction Model with Time dependant features
I got a very long sequence of binary items (0 or 1). Each item is associated to a timestamp. For example :
...
5
votes
1
answer
667
views
What are disadvantages of "Sequential analysis"
Sequential analysis involves performing sequential interim analysis till results are significant or till a maximum number of interim analyses is reached.
Sequential analysis sounds appealing ...
1
vote
0
answers
31
views
Average Sample Number using Page's CUSUM procedure?
I am trying reproduce the Table 1 results from the page 12 using the equation given the page 13.To access the journal article please click https://arxiv.org/pdf/math/0605322.pdf. The corresponding ...
3
votes
1
answer
165
views
Does an interim sample size re-estimation increase type 1 error if based on the overall event rate?
Suppose that a sample size is to be calculated for a trial: a presumed effect size summarizes the effect of intervention (relative risk RR $\exp(\theta_1)$), and background gives the rate of events in ...
2
votes
0
answers
55
views
Mixed-effects modelling in a sequential analysis design
I'm planning a study and am considering going with a sequential analysis design (as described here: https://psyarxiv.com/9yegd/download?format=pdf). This design allows one to take multiple looks at ...
0
votes
0
answers
26
views
How to methodologically show that a given 'time-series/sequential' data is not really sequential?
I have an apparent time-series/sequencial (supervised: multi-class classification problem) dataset with each data-point time-stamped. However, some domain intuition tells me that the data is static ...
-1
votes
1
answer
536
views
Reverse the shuffling phase
It is very convenient to have sklearn library to split and shuffle the data with
...
4
votes
1
answer
117
views
Can the Bayesian but not the frequentist "just add more observations"?
Since the frequentist's p-values are uniformly distributed under the null hypothesis, it is a highly problematic practice to add more and more data to your sample until you find a significant result. ...
1
vote
0
answers
14
views
Sequential mining technique for categorical events?
I'm facing a problem where I have a short (2-8) sequence of discrete events, of variable length, and a final state. Ex.: cold, warm, warm -> soft clothing.
I would like to determine what are the main ...
2
votes
0
answers
77
views
Predicting elections with a random sample
If I have a population of exactly 100.000 voters, and I know they all voted in either candidate A, B, C, D or E. (e.g where the results of a random sample of 100 voters are A = 40%, B = 30%, C = 10%, ...
1
vote
1
answer
49
views
Applying Sequential Probability Ratio Test (SPRT) on a Vector Signal?
I am trying to apply SPRT on a vector signal (of residual) for detecting faults occurring in the sensors' measurement of an aircraft. The vector signal contains measurements from five different ...
6
votes
1
answer
278
views
Stopping rules in group sequential designs
I have read that in group sequential designs, having a maximum of $k$ stops, for the $j$-th stop $j < k$ there are 4 bounds chosen $a_j,b_j, c_j, d_j$ such that the three decisions are considered. ...
2
votes
0
answers
30
views
power calculation for two stage binomial type model
This is a power-type calculation for a Bernoulli/binomial question in two stages.
Suppose you are planning an experiment which starts with a test for an event on $N$ experimental units. The event ...
12
votes
5
answers
2k
views
Determining sample size before starting an experiment or run the experiment indefinitely?
I studied statistics years ago and have forgotten it all so these may seem like general conceptual questions than anything specific but here is my issue.
I work for an e-commerce website as a UX ...
4
votes
1
answer
2k
views
How many samples do I need to estimate a gaussian distribution?
Background: I'm comparing the performance of two algorithms. The data is noisy so the results tend to form a nice normal distribution.
I'd like to estimate the mean with a high confidence: up to 1% ...
3
votes
0
answers
28
views
Advice for conducting the same experiment several times?
I'm very new to Bayesian statistics, but I've found myself in a situation where they might be able to help. I'll be running the same experiment repeatedly and analyzing the data every few months. The ...
2
votes
0
answers
422
views
Understanding interim analysis and alpha spending functions
I have to perform an A/B test with interim analysis and found some literature about methods designed to preserve the global Type I error rate by adjusting the $\alpha$ level for each peeking.
...
1
vote
0
answers
76
views
Convergence of a sequence in finite number of steps
Here is the setup of my problem. It is a sequential problem and there are two possible actions A and B. Now, when either action $A$ or $B$ is taken at the $j$th time point, we observe some outcome say ...
2
votes
0
answers
49
views
Optimal denoising of a few features for latent variable modeling
I am fitting latent variable models to noisy, high-dimensional data. The dimension is 15,000 to 20,000; there are 2,000 observations. There is a certain latent structure that I observe -- essentially, ...
1
vote
1
answer
917
views
Bayesian updating with discrete priors + possibly unknown classes
I'm following along with some lecture notes on Bayesian updating with discrete priors. They give an example problem to illustrate some of these concepts, which I briefly restate here:
Someone tells ...
1
vote
1
answer
391
views
sample size calculation for one-sample log rank test with interims (ie group sequential)
I wonder if there is any advice for performing a sample size calculation for survival data using the one-sample log rank test with interims (ie group sequential) - ie with O'Brien-Fleming or some ...