Science topics: MathematicsStatistics
Science topic
Statistics - Science topic
Statistical theory and its application.
Questions related to Statistics
Hi all!
I've run both LFQ and TMT 18-plex proteomics on the same protein extracts.
My experiment consists of two study conditions, and 8 biological replicates.
After digesting my protein extractions, I ran half of the peptide preparation using DDA with four technical replicates, and the other half I TMT tagged (18-plex, two reference channels, one mixture) fractionated, and ran using an SPS MS3 method on the Fusion Lumos.
I've done the searches in PD2.4, and summarised the results with `MSstats` and `MSstatsTMT`.
I'm currently working on how to deal with two different datasets of the same experiment, the original plan was to use the LFQ dataset for the improved coverage, and the TMT dataset for improved quantification.
One thing I've noticed is that while the TMT dataset has significantly better adjusted p-values, the fold changes are less pronounced than the LFQ dataset, meaning that quite a few proteins fail the biological significance thresholds. See the attached volcano plots (vertical dotted lines represent 0.58 log2 FC, horizontal 0.05 adjusted p-value). The scales are not consistent between the plots sorry!
I'm aware that MS2 TMT methods have an issue with reporter ion compression blunting fold change values, and was hoping that it would be less of an issue with my MS3 method. Is there a correction for this, or does this reflect a lack of dramatic fold-change in my biology?
Any other tips for integrating LFQ and TMT data would also be appreciated!
Sam


I am currently conducting a study using a photothrombotic stroke model in C57BL6 mice and measuring motor function outcome following strokes to determine if a pharmacological treatment can help improve their motor recovery. To measure motor recovery, I am using the tapered beam task and the grid walk task. Both of these tasks measure the number of errors that the mice make during trials. One thing that I've noticed is that a handful of the mice in the placebo group (no pharmacological treatment, just saline) are unable to complete the tasks on the first day of behavior due to the severity of the injury and the lack of treatment.
As such, I'm wondering if there is a standard way to handle missing data that is a result of severe injuries and is important for accurately reflecting differences between my groups. The methods that I can think of would either be filling with the mean for the group, filling with the highest number of errors of the group (e.g. the worst recorded score was 93 errors in the placebo group, presumably the mice unable to complete the task have more severe strokes and should receive the max number of errors observed), or multiple imputation using the MICE package in R. My understanding is that multiple imputation is the standard for filling in data that is not missing at random, but I want to ensure that is true in this scenario as well. Any citations (especially those specific to animal models) to support these methods would be greatly appreciated as well.
𝑃 ( 𝑋 ≥ 𝑘 − 1 ) for X∼Binomial(n−1,p).
P(X≥k) for X∼Binomial(n,p)

I have seen many comments implying if a newly developed scale has a solid model background, EFA can (or better, should) be skipped. In a cognitive scale that I have recently developed, I had a clear design on my items, based on the previous theory. However, after administrating it to my study population, I ran a WLSMV CFA with two first-order factors and saw that some items (out of a total of 50) have weak (<0.30) or very weak (<0.10) loadings and possible cross-loadings.
My fit indices improved to an excellent range after deleting some of the lowest-loading items. Even after that, I have items with factor loadings of ~0.20. I have good reliability when they stay. And they don't look bad, theoretically. After pruning them to have a minimum loading of 0.3, not only my already good fit indices don't improve much, but my reliability gets lower. And I lose a good chunk of items. You don't want to assess cognitive skills with 15 items since almost all batteries have 30-40 items minimum. Should I keep them?
Also, some of the items with ceiling effect (98% correct responses) stay in the CFA model with good loadings. Should I keep them?
There are clear guidelines on item-deleting strategies for EFA. What about CFA?
What’s the most common programming paradigm of no-code platforms? Why?
I am at the end of conducting a large systematic review and meta-analysis. I have experience of meta-analysis and have attempted to meta-analyse the studies myself, but I am not happy with my method. The problem is that almost all the studies are crossover studies and I am not sure how to analyse them correctly. I have consulted the Cochrane Handbook, and it seems to suggest a paired analysis is best, but I do not have the expertise to do this - https://training.cochrane.org/handbook/current/chapter-23#section-23-2-6
I am seeking a statistician familiar with meta-analysis to consult with, and if possible, undertake the meta-analysis. There are only two authors on this paper (me and a colleague), so you would either be second or last author. We aim to publish in a Q1 or Q2 journal, and from my own analysis I can see we have very interesting results.
Please let me know if you are interested.
Hello network,
Is there anyone who could help me with my empirical research - especially with the statistical elaboration - on the topic of entrepreneurial intention and business succession in German SMEs, or who has experience with the preparation of Structural Equation Modeling?
Please feel free to send me a private message.
Thank you and best regards
Julia
Why are people of color supporting unchecked white power?
I plan to perform a t-test between risk index and leakage values to know if there is a significant difference between leakage values of low, medium, and high risk pipes. However, the data I have on leakage and risk index was measured using the same raw data (e..g. pipe age, pipe material, pipe diameter, pressure)
Hi all, I wanted to see whether extrinsic or intrinsic motivation affects the frequency of cosplaying. The questionnaire for motivation has 4 subscales, in which I mark them as E1 (social), E2 (social avoidance), I1 (intellectual), I2 (competency mastery). The frequency was measured as whether participants have cosplayed (meaning dressing up as a character at any event) once, twice, or more than thrice. Logically, intrinsic and extrinsic motivation influence each other, but I thought that it would be interesting to see whether there are any significant differences between each sub for this particular sample. In other words, it's like finding out, for example, which one contributes more to the behaviour of cosplaying; is it because of increasing skill mastery or a sense of belonging?
IV: types of motivation with 2 levels (intrinsic, extrinsic). possibly it's counted as 4 levels if I regard each subconstruct (social, social avoidance, intellectual, competency master) as 1 level.
DV: frequency of cosplay
My current hypotheses are kept simple for now:
H1: There is a relationship between extrinsic motivation and the frequency of cosplaying.
H2: There is a relationship between intrinsic motivation and the frequency of cosplaying.
I'm getting confused after all the research for a suitable test. For now, I'm clear that I have an ordinal IV and DV. However, I'm unsure where should I begin looking if I were to find the differences in effect or predicting which type (E1, E2, I1 or I2) of motivation subconstruct would contribute to the frequency of behaviour. I'm looking into ordinal regression, linear regression, ANOVA ... but none of them seem to be suitable. Originally, I thought of doing Pearson correlation since that seems to be an obvious choice, but I'd like to explore more about the 4 subconstructs. Unless I'm not looking at my whole picture correctly, I would really appreciate input and assistance. I'm more than happy to give more details about the research to help in this inquiry. Thank you!
Factor loading is more than 0.92 also ave and cr are more than 0.95? Is this acceptable? If it's not i tried to delete items but the more i delete the more factor loading is higher! how can i treat it?
I have 6 ecosystems, 3 of which are substrate A and the other 3 are substrate B. each ecosystem has about 10 species. I have calculated a simpsons value for each ecosystem and a simpsons value for each substrate. I would like to statistically compare the two index values of substrate A and B, is this possible in any way? Since I would like to statistically compare the biodiversity between the two substrates, what is the best way to go about this?
I have six ecosystems in two substrate categories (Triplicates essentially). I have determined shannon wiener index values for each ecosystem and also for the two categories separately. I have done this for two separate sets of data that were sampled in two separate years. Is it possible to statistically compare the development of the biodiversity between each of the categories i.e., the development of biodiveristy in ecosystem 1 between the two years, using the shannon wiener values somehow? Are there any other tests that could work? I am aware of the hutcheson t test however, some of my data is not normally distributed.
I would really appreciate some help!
I have six ecosystems in two substrate categories (Triplicates essentially). I have determined shannon wiener index values for each ecosystem and also for the two categories separately. I have done this for two separate sets of data that were sampled in two separate years. Is it possible to statistically compare the development of the biodiversity between each of the categories i.e., the development of biodiveristy in ecosystem 1 between the two years, using the shannon wiener values somehow? Are there any other tests that could work? I am aware of the hutcheson t test however, some of my data is not normally distributed.
I would really appreciate some help!
Is it very literally subbing in shannon wiener index values instead of species abundances?
So my student have a question that i cannot answer as well. She analyzing the effect of ICT toward labor productivity using 8 years data panel using 4 independent variables with EVIEWS 13. Frankly i quite surprised that the R-squared value on her results is 0.94 with only 2 significance variables. Theoretically, in simple regression model with higher value of R-squared most likely indicates bad and have statistics problems. Recently, i asked her to calculate the data using STATA and the results shows that only have 0,51 R-Square with exact similar coefficient.
I've search some articles about it and says that eviews might be wrong, and some says that STATA is wrong. Can someone explain what should i do and which software have to use?
note:
1. Some articles says to using areg command in stata to find similar value as eviews, but i quite doubt because areg is using for categorical regression in stata and its not quite fit in panel regression model.
2. Some says that eviews is wrong calculation.


Hello,
I discovered that a compound that I use, which get integrated into RNA, might have an unspecific impact on mRNA stability of a gene of interest.
To confirm that, I pretreated my cells for 2h with this compound to let it be incorporated into RNA, then did a timecourse with Actinomycin D to block transcription and observe my target of interest mRNA stability through time.
After qPCR, I have a list of RQ values, all calculated using an untreated timepoint 0 control as a reference. My data are the following : DMSO, Compound, Actinomycin D, Compound + Actinomycin D at 1, 2, 3 and 4 hours + the timepoint 0 control. Experiment was performed 4 times.
Once plotted, the results give me 4 lines, each of them representing the impact of one treatment through time. To prove that my compound impacts mRNA stability, I need to prove that the line with the Compound + Actinomycin D is statistically significantly lower compared to Actionmycin D alone.
Could you help me to select the best statistical test to use for this question ?
So far, here are the other strategies I tried:
1 - I analyzed the qPCR data using each DMSO condition as a RQ reference for their respective timepoint, giving me a histogram comparing all treatments together at each timepoint and performed a two-way ANOVA on them. If I'm correct, this strategy assesses at which timepoint, the treatments are different from each other. However, I would like now to analyse the data globally, and not separately, timepoint by timepoint.
2 - I followed the data analysis section of this paper :
Which, if I understand it correctly, calculates how well each of my curves will follow a decay model to calculate a decay rate. However, this is not exactly the answer I want and moreover, Prism gave me the following answer :
One phase decay - Least squares fit
Prism has identified at least one unstable parameter.This suggests that your data may be incomplete or don’t fully describe the selected model.Because of this, confidence intervals for all parameters should be interpreted with caution.
For at least one parameter, Prism was able to find a best-fit value, but was unable to calculate a complete confidence interval. This best-fit value should be interpreted with caution.
As I think this other strategy doesn't fit my needs and my data don't seem to be adequate for it, I do not plan on doing more on this second strategy.
Thank you very much for your help.
Assuming this is my hypothetical data set (attached figure), in which the thickness of a structure was evaluated in the defined positions (1-3) in 2 groups (control and treated). I emphasize that the structure normally increases and decreases in thickness from position 1 to 3. I would also like to point out that each position has data from 2 individuals (samples).
I would like to check if there is a statistical difference in the distribution of points (thickness) depending on the position. Suggestions were to use the 2-sample Kolmogorov-Smirnov test.
However, my data are not absolutely continuous, considering that the position of the measurement in this case matters (and the test ignores this factor, just ordering all values from smallest to largest and computing the statistics).
In this case, is the 2-sample Komogorov-Smirnov test misleading? Is there any other type of statistical analysis that could be performed in this case?
Thanks in advance!

"DNA is SO unpredictable that they are either fractals or something less predictable, thus a gene is never known to manifest into a trait, debunking hereditarianism and vindicating CRT" (Ohnemus 2024).
Dear colleagues,
I would like to ask whether it is possible to compare the quality of different models based on the same data but containing different number of variables using IRT analysis, namely Log likelihood, AIC and BIC statistics?
Specifically, I have a model with 36 items and I am gradually eliminating some problem variables and I want to compare the overall quality of the model between each other based on the above statistics?
Is this procedure possible?
Thank you for your answer.
Dear colleagues,
I would like to ask for your advice on testing the criterion-related validity of the measuring instrument. It is common practice to test this type of validity by correlation with other relevant variables. However, I received a comment from a reviewer that if I calculate only Pearson correlation, the measurement error is not taken into account and the correlation is underestimated.
He said I should use reliability-corrected correlations or report the correlation by fitting an SEM model where the three factors correlated with the external variables (my measurement instrument is a simple structure with three correlated factors).
Could I ask your advice on how to calculate this? Personally, I do not know how I should proceed. Alternatively, what is your opinion?
Thank you very much.
I have a mixed effect model, with two random effect variables. I wanted to rank the relative importance of the variables. The relimpo package doesn't work for mixed effect model. I am interested in the fixed effect variables anyway so will it be okay if I only take the fixed variables and use relimp? Or use weighted Akaike for synthetic models with alternatively missing the variables?
which one is more acceptable?
Hello all,
I am running into a problem I have not encountered before with my mediation analyses. I am running a simple mediation X > M > Y in R.
Generally, I concur that the total effect does not have to be significant for there to be a mediation effect, and in the case I am describing, this would be a logical occurence, since the effects of path a and b are both significant and respectively are -.142 and .140, thus resulting in a 'null-effect' for the total effect.
However, my 'c path X > Y is not 'non-significant' as I would expect, rather, the regression does not fit (see below) :
(Residual standard error: 0.281 on 196 degrees of freedom
Multiple R-squared: 0.005521, Adjusted R-squared: 0.0004468
F-statistic: 1.088 on 1 and 196 DF, p-value: 0.2982).
Usually I would say you cannot interpret models that do not fit, and since this path is part of my model, I hesitate to interpret the mediation at all. However, the other paths do fit and are significant. Could the non-fitting also be a result of the paths cancelling one another?
Note: I am running bootstrapped results for the indirect effects, but the code does utilize the 'total effect' path, which does not fit on its own, therefore I am concerned.
Note 2: I am working with a clinical sample, therefore the samplesize is not as great as I'd like group 1: 119; group2: 79 (N = 198).
Please let me know if additional information is needed and thank you in advance!
Dear colleagues
Could you tell me please,how is it possible to consruct boxplot from dataframe in rstuio
df9 <- data.frame(Kmeans= c(1,0.45,0.52,0.54,0.34,0.39,0.57,0.72,0.48,0.29,0.78,0.48,0.59),hdbscan= c(0.64,1,0.32,0.28,0.33,0.56,0.71,0.56,0.33,0.19,0.53,0.45,0.39),sectralpam=c(0.64,0.31,1,0.48,0.24,0.32,0.52,0.66,0.32,0.44,0.28,0.25,0.47),fanny=c(0.64,0.31,0.38,1,0.44,0.33,0.48,0.73,0.55,0.51,0.32,0.39,0.57),FKM=c(0.64,0.31,0.38,0.75,1,0.26,0.55,0.44,0.71,0.38,0.39,0.52,0.53), FKMnoise=c(0.64,0.31,0.38,0.75,0.28,1,0.42,0.45,0.62,0.31,0.25,0.66,0.67), Mclust=c(0.64,0.31,0.38,0.75,0.28,0.46,1,0.36,0.31,0.42,0.47,0.66,0.53), PAM=c(0.64,0.31,0.37,0.75,0.28,0.46,0.58,1,0.73,0.43,0.39,0.26,0.41) ,
AGNES=c(0.64,0.31,0.37,0.75,0.28,0.46,0.58,0.55,1,0.31,0.48,0.79,0.31), Diana=c(0.64,0.31,0.37,0.75,0.28,0.46,0.58,0.55,0.42,1,0.67,0.51,0.43),
zones2=c(0.64,0.31,0.37,0.75,0.28,0.46,0.58,0.55,0.42,0.45,1,0.69,0.35),
zones3=c(0.64,0.31,0.37,0.75,0.28,0.46,0.58,0.55,0.42,0.45,0.59,1,0.41),
gsa=c(0.64,0.31,0.37,0.75,0.28,0.46,0.58,0.55,0.42,0.45,0.59,0.36,1), method=c("kmeans", "hdbscan", "spectralpam", "fanny", "FKM","FKMnoise", "Mclust", "PAM", "AGNES", "DIANA","zones2","zones3","gsa"))
head(df9)
df9 <- df9 %>% mutate(across(everything(), ~as.numeric(as.character(.))))
Thank you ery much
In the domain of clinical research, where the stakes are as high as the complexities of the data, a new statistical aid emerges: bayer: https://github.com/cccnrc/bayer
This R package is not just an advancement in analytics - it’s a revolution in how researchers can approach data, infer significance, and derive conclusions
What Makes `Bayer` Stand Out?
At its heart, bayer is about making Bayesian analysis robust yet accessible. Born from the powerful synergy with the wonderful brms::brm() function, it simplifies the complex, making the potent Bayesian methods a tool for every researcher’s arsenal.
Streamlined Workflow
bayer offers a seamless experience, from model specification to result interpretation, ensuring that researchers can focus on the science, not the syntax.
Rich Visual Insights
Understanding the impact of variables is no longer a trudge through tables. bayer brings you rich visualizations, like the one above, providing a clear and intuitive understanding of posterior distributions and trace plots.
Big Insights
Clinical trials, especially in rare diseases, often grapple with small sample sizes. `Bayer` rises to the challenge, effectively leveraging prior knowledge to bring out the significance that other methods miss.
Prior Knowledge as a Pillar
Every study builds on the shoulders of giants. `Bayer` respects this, allowing the integration of existing expertise and findings to refine models and enhance the precision of predictions.
From Zero to Bayesian Hero
The bayer package ensures that installation and application are as straightforward as possible. With just a few lines of R code, you’re on your way from data to decision:
# Installation
devtools::install_github(“cccnrc/bayer”)# Example Usage: Bayesian Logistic Regression
library(bayer)
model_logistic <- bayer_logistic( data = mtcars, outcome = ‘am’, covariates = c( ‘mpg’, ‘cyl’, ‘vs’, ‘carb’ ) )
You then have plenty of functions to further analyze you model, take a look at bayer
Analytics with An Edge
bayer isn’t just a tool; it’s your research partner. It opens the door to advanced analyses like IPTW, ensuring that the effects you measure are the effects that matter. With bayer, your insights are no longer just a hypothesis — they’re a narrative grounded in data and powered by Bayesian precision.
Join the Brigade
bayer is open-source and community-driven. Whether you’re contributing code, documentation, or discussions, your insights are invaluable. Together, we can push the boundaries of what’s possible in clinical research.
Try bayer Now
Embark on your journey to clearer, more accurate Bayesian analysis. Install `bayer`, explore its capabilities, and join a growing community dedicated to the advancement of clinical research.
bayer is more than a package — it’s a promise that every researcher can harness the full potential of their data.
Explore bayer today and transform your data into decisions that drive the future of clinical research: bayer - https://github.com/cccnrc/bayer

What may be a good, strong and convincing example demonstrating the power of copulas by uncovering some not obvious statistical dependencies?
I am especially interested in the example contrasting copula vs a simple calculation of a correlation coefficient for the original distributions.
Something like this - the (properly normalized) correlation coefficient of components of a bivariate distribution does not suggest a strong statistical dependence between them, but the copula distribution of these two components shows a clear dependence between them (possibly manifested in the value of a correlation coefficient calculated for the copula distribution?). Or the opposite - the correlation coefficient of the original bivariate distribution suggests strong dependence, but its copula shows that the statistical dependence is "weak", or just absent.
Mostly interested in an example described in terms of formulae (so that the samples could be generated, e.g. in MATLAB), but if somebody can point to the specific pre-generated bivariate distribution dataset (or its plots), that will work too.
Thank you!
I want to estimate the half-life value for the virus as a function of strain and concentration, and as a continuous function of temperature.
Could anybody tell me, how to calculate the half-life value in R programming?
I have attached a CSV file of the data
I am studying leadership style's impact on job satisfaction. in the data collection instrument, there are 13 questions on leadership style divided into a couple of leadership styles. on the other hand, there are only four questions for job satisfaction. how do i run correlational tests on these variables? What values do i select to analyze in Excel?
I explain here the connection between the pre-scientific Law of Universal Causality and all sorts of statistical explanations in physical sciences. The way it takes may look strange, but it will be interesting enough to consider.
To repeat in short what is already said a few times: by all possible assumptions, to exist (which is To Be with respect to Reality-in-total) is non-vacuous. Hence, any existent must have Extension, have finite-content parts. These parts, by the only other possible assumption, must yield impacts on other parts both external and internal. This is Change.
These impacts are always finite in the content and measured extents. The measured extents of Extension and Change are space and time. Without measurements we cannot speak of space and time as existing or as pertaining to existents. What pertain to all existents as most essential are Extension and Change. Existence in Extension and Change means that finitely extended objects give origin to finite impacts. This is Causality. Every existent is Extension-Change-wise existent, and hence everything is causal.
As pertinents to existents, Extension and Change are the most applicable qualities / universals of the group of all entities, i.e., Reality-in-total, because they belong to all that exist. Since Extension and Change are not primarily in our minds, let us call them as ontological universals. As is clear now, Extension and Change are the widest possible and most general ontological universals. All universals are pure qualities. All qualities other than ontological universals are mixtures of pure qualities.
There are physical-ontological universals / qualities that are not as universal as Extension and Change. ‘Colouredness’ / ‘being coloured’, ‘redness’, ‘unity’ / ‘being a unit’, ‘being malleable’, ‘being rigid’, etc. are also pure qualities. These are pertinents not merely of one existent process. They belong to many. These many are a group of existent processes of one kind, based on the one classification quality. Such groups of Extension-Change-wise existent entities are termed natural kinds.
Ontological universals can be reflected in minds too, but in very meagre ways, not always, and not always to the same extent of correspondence with ontological universals, because they are primarily in existent processes. A direct reflection is impossible. The many individuals who get them reflected meagrely formulate them differently.
The supposed common core of ontological universals in minds is a pure notion, but they are mere notions idealized by minds. These ideals are also not inherited of the pertinent ontological universals of all relevant existent things, but at least by way of absorption from some existents, in whatever manner of correspondence with ontological universals. I call them connotative universals, because they are the pure aspects of the conceptual activity of noting objectual processes together.
In brains connotative universals can show themselves only as a mixture of the relevant connotative universals and the relevant brain elements. Please note that this is not a brain-scientific statement. It is the best imaginable philosophical common-sense on the brain-scientific aspect of the formation of connotative universals, and hence it is acceptable to all brain scientists. In brains there are processes that define such activities. But it needs only to be accepted that these processes too are basically of Extension-Change-wise existence, and hence are causal in all senses.
Connotatives are just representations of all kinds of ontological universals. Connotatives are concatenated in various ways in connection with brain elements – in every case highly conceptually and symbolically. These concatenations of connotatives among themselves are imaginations, emotions, reflections, theories, etc., as considered exclusively in the mind.
Note here also that the lack of exact correspondence between ontological and connotative universals is what makes ALL our statements essentially statistical and non-exact at the formation of premises and at the jump from premises into conclusions. The statistical aspect here is part of the process of formation, by brains, of connotatives from ontological universals. This is the case in every part of imaginations, emotions, reflections, theories, etc., even when statistical measurements are not actually being made part of the inquiry as a matter of mentally guided linguistic and mathematical procedures.
Further, connotative universals are formulated in words expressed as terms, connected with connectives of processes, and concatenated in statements. These are the results of the symbolic functioning of various languages including mathematics. These are called denotative universals and their concatenations. All symbolic activities function at this level.
Now coming to statistics as an applied expression of mathematics. It is nothing but denotative universals concatenated in a quantitatively qualitative manner. Even here there is a lot of lack of exactness, which are known as uncertainty, randomness, etc. Pay attention to the fact that language, mathematics, and its statistical part work at the level of denotative universals and their concatenations. These are naturally derived from the conglomerations of ontological universals via concatenations of connotatives and then translated with further uncertainties unto denotative concatenations.
Causation works at the level of the conglomerations of ontological universals, which are in existent things themselves. That is, statistical connections appear not at the ontological level, but at the denotative level. When I say that this laptop is in front of me, there is a directness of acceptance of images from the ontological universals and their conglomerations into the connotative realm of connotations and from there into the denotative realm of connotations. But in roundabout conclusions regarding causal processes at the physical-ontological level into the statistical level, the amount or extent of directness of judgement is very much lacking.
What is the specific importance of a bachelor’s degree in the hiring process?
Why parsimoniously does fertility negatively correlate with socioeconomic status? How?
Hi,
I am hoping to get some help on what type of statistical test to run to validate my data. I have run 2 ELISAs with the same samples for each test. I did perform a Mann-Whitney U-test to compare the groups, and my results were good.
However, my PI wants me to also run a statistical test to determine that there wasn't any significant difference in the measurement of each sample between the runs. He wants to know that my results are concordant/reproducible.
I am trying to compare each sample individually, and since I don't have 3 data points, I can't run an ANOVA. What types of statistical tests will give me that information? Also, is there a test that will run all the samples simultaneously but only compare across the same sample.
For example, if my data looked like this.
A: 5, 5.7
B: 6, 8
C: 10, 20
I need a test to determine if there is any significant difference between the values for samples A, B, and C separately and not compare the group variance between A-C.
Hello everyone,
I am currently undertaking a research project that aims to assess the effectiveness of an intervention program. However, I am encountering difficulties in locating suitable resources for my study.
Specifically, I am in search of papers and tutorials on multivariate multigroup latent change modelling. My research involves evaluating the impact of the intervention program in the absence of a control group, while also investigating the influence of pre-test scores on subsequent changes. Additionally, I am keen to explore how the scores differ across various demographic groups, such as age, gender, and knowledge level (all measured as categorical variables).
Although I have come across several resources on univariate/bivariate latent change modelling with more than three time points, I have been unable to find papers that specifically address my requirements—namely, studies focusing on two time points, multiple latent variables (n >= 3), and multiple indicators for each latent variable (n >= 2).
I would greatly appreciate your assistance and guidance in recommending any relevant papers, tutorials, or alternative resources that pertain to my research objectives.
Best,
V. P.
I want to examine the relationship between school grades and self-esteem and was planning to do a linear regression analysis.
Here's where my Problem is. I have three more variables: socioeconomic status, age and sex. I wanted to treat those as moderation variables, but I'm not sure if that's the best solution. Maybe a multiple regression analysis would be enough? Or should I control those variables?
Also if I'd go for a moderation analysis, how'd I go about analysing with SPSS? I can find a lot of videos about moderation analysis, but I can't seem to find cases with more than one moderator.
I've researched a lot already but can't seem to find an answer. Also my statistic skills aren't the best so maybe that's why.
I'd be really thankful for your input!
I recently had a strange question from one of the non-statistician asking on confidence interval. His way of understanding is that all the sample values that was used to calculate the confidence interval should be within that interval. I have tried to answer him the best, but couldn't convince him in any way. Is there any best way to explain why it need not be, and the purpose is not the way he understands. How would you handle this question?
Thanks in advance.
Respectfully, across reincarnation belief and scientific materialism, why is considering the individual self, as an illusion, a commonality? 1)
Dear all,
I am sharing the model below that illustrates the connection between attitudes, intentions, and behavior, moderated by prior knowledge and personal impact perceptions. I am seeking your input on the preferred testing approach, as I've come across information suggesting one may be more favorable than the other in specific scenarios.
Version 1 - Step-by-Step Testing
Step 1: Test the relationship between attitudes and intentions, moderated by prior knowledge and personal impact perceptions.
Step 2: Test the relationship between intentions and behavior, moderated by prior knowledge and personal impact perceptions.
Step 3: Examine the regression between intentions and behavior.
Version 2 - Structural Equation Modeling (SEM)
Conduct SEM with all variables considered together.
I appreciate your insights on which version might be more suitable and under what circumstances. Your help is invaluable!
Regards,
Ilia

ResearchGate does a pretty good job of tracking publication analytics such as reads and citations over time. The recommendations feature can also be an interesting indicator for a publication's resonance with the scholarly community.
This progress allow for ideas to be developed about how to make the analytics features even better in the future. Here are some ideas I have been thinking about:
- something equivalent to Altmetric that tracks social media mentions across multiple platforms and mentions in news articles, conference proceedings, etc.
- more longitudinal data for individual publications by month and year
- the ability to compare the performance of one's own publications, with perhaps a way to rank them in analytic reports by reads, citations, etc.
- More specific analytics to allow for comparisons within and between departments on an individual and collective basis, which can be sorted by discipline, field, etc.
Are there any additional analytics features that you would like to see on ResearchGate?
Meta-analyses and systematic reviews seem the shortcut to academic success as they usually have a better chance of getting published in accredited journals, be read more, and bring home a lot of citations. Interestingly enough, apart from being time-consuming, they are very easy; they are actually nothing but carefully followed protocols of online data collection and statistical analysis, if any.
The point is that most of this can be easily done (at least in theory) by a simple computer algorithm. A combination of if/thenstatements would simply allow the software to decide on the statistical parameters to be used, not to mention more advanced approaches that can be available to expert systems.
The only part needing a much more advanced algorithm like a very good artificial intelligence is the part that is supposed to search the articles, read them, accurately understand them, include/exclude them accordingly, and extract data from them. It seems that today’s level of AI is becoming more and more sufficient for this purpose. AI can now easily read papers and understand them quite accurately. So AI programs that can either do the whole meta-analysis themselves, or do the heavy lifting and let the human check and polish/correct the final results are on the rise. All needed would be the topic of the meta-analysis. The rest is done automatically or semi-automatically.
We can even have search engines that actively monitor academic literature, and simply generate the end results (i.e., forest plots, effect sizes, risk of bias assessments, result interpretations, etc.), as if it is some very easily done “search result”. Humans then can get back to doing more difficult research instead of putting time on searching and doing statistical analyses and writing the final meta-analysis paper. At least, such search engines can give a pretty good initial draft for humans to check and polish them.
When we ask a medical question from a search engine, it will not only give us a summary of relevant results (the way the currently available LLM chatbots do) but also will it calculate and produce an initial meta-analysis for us based on the available scientific literature. It will also warn the reader that the results are generated by AI and should not be deeply trusted, but can be used as a rough guess. This is of course needed until the accuracy of generative AI surpasses that of humans.
It just needs some enthusiasts with enough free time and resources on their hands to train some available open-source, open-parameter LLMs to do this specific task. Maybe even big players are currently working on this concept behind the scene to optimize their propriety LLMs for meta-analysis generation.
Any thoughts would be most welcome.
Vahid Rakhshan
hi, i'm currently writing my psychology dissertation where i am investigating "how child-oriented perfectionism relates to behavioural intentions and attitudes towards children in a chaotic versus calm virtual reality environment".
therefore i have 3 predictor variables/independent variables: calm environment, chaotic environment and child-oriented perfectionism
my outcome/dependent variables are: behavioural intentions and attitudes towards children.
my hypotheses are:
- participants will have more negative behavioural intentions and attitudes towards children in the chaotic environment than in the calm environment.
- these differences (highlighted above) will be magnified in participants high in child-oriented perfectionism compared to participants low in child oriented perfectionism.
i used a questionnaire measuring child-oriented perfectionism which will calculate a score. then participants watched the calm environment video and then answered the behavioural intentions and attitudes towards children questionnaires in relation to the children shown in the calm environment video. participants then watched the chaotic environment video and then answered the behavioural intentions and attitudes towards children questionnaire in relation to the children in the chaotic environment video.
i am unsure whether to use a multiple linear regression or repeated measures anova with a continuous moderator (child-oriented perfectionism) to answer my research question and hypotheses. please please can someone help!
RuntimeWarning: Precision loss occurred in moment calculation due to catastrophic cancellation. This occurs when the data are nearly identical. Results may be unreliable.
res = hypotest_fun_out(*samples, **kwds)
Above warning occured in python. Firstly, the dataset was normalised and then while performing the t-test this warning appeared, though the output was displayed. Kindly suggest some methods to avoid this warning.
I am somewhat Hegelian because I do not believe in martyrdom, and or dying on a hill, and usually the popular, and or traditional, opinion has a deeper less obvious reason.
Can anyone here me with one biostatistics question. It is about finding the sample size from power analysis. I have the variables. Just need an assistance with the calculations.
As a Computer Science student inexperienced in statistics, I'm looking for some advice on selecting the appropriate statistical test for my dataset.
My data, derived from brain scans, is structured into columns: subject, channels, freqbands, measures, value, and group. It involves recording multiple channels (electrodes) per patient, dividing the signal into various frequency bands (freqbands), and calculating measures like Shannon entropy for each. So each signal gets broken down to one data point. This results in 1425 data points per subject (19 channels x 5 freqbands x 15 measures), totalling around 170 subjects.
I aim to determine if there's a significant difference in values (linked to specific channel, freqband, and measure combinations) between two groups. Additionally, I'm interested in identifying any significant differences at the channel, measure or freqband level.
What would be a suitable statistical test for this scenario?
Thanks in advance for any help!
Has anyone gone through the Wohler's report_2023 yet? Its pros and cons? What are the ways to obtain its e-copy? Its subscription is very costly for a normal researcher (around 750 USD per user). Any alternatives to get similar kind of data as that of the report?
I'm excited to speak at this FREE conference for anyone interested in statistics in clinical research. 👇🏼👇🏼
The Effective Statistician conference features a lineup of scholars and practitioners who will speak about professional & technical issues affecting statisticians in the workplace.
I'll be giving a gentle introduction to structural equation modeling! I hope to see you there.
Sign up here:
Is it correct to choose the principal components method in order to show the relationship of species with biotopes?
My answer: Yes, in order to interpret history, disincentives are the most rigorous guide. How?: Due to the many assumptions of inductive logic, deductive logic is more rigorous. Throughout history, incentives are less rigorous because no entity(besides God) is completely rational and or self-interested, thus what incentivizes an act is less rigorous then what disincentivizes the same action. And, as a heuristic, all entities(besides God) have a finite existence before their energy(eternal consciousness) goes to the afterlife( paraphrased from these sources : 1)
2) )
, thus interpretation through disincentives is more rigorous than interpreting through incentives.
Who agrees life is more about preventing tragedies than performing miracles? I welcome elaborations.
If you're using a number such as a statistic from a reference study you want to cite, should you write the number with the confidence interval? And how to effectively prevent plagiarism when dealing with numbers?
Thank you!
Are people more likely to mix up words if they are fluent in more languages? How? Why?
Hi!
This might be a bit of a stupid question, but I am currently writing my master thesis. One of the things I am doing is a factor analysis on a scale developed in Canada. This scale has only been validated on the Canadian workforce (the developers have one time done a exploratory factor analysis and two times done a confirmatory factor analysis). I am doing an exploratory and a confirmatory factor analysis in the Norwegian workforce to see what factor structure I would find here, and if it is the same as in Canada. As this this only one of three things I am doing in my masters I have hypothesis for all the other findings, so my supervisor would like me to have hypothesis for the factor structure as well. Whenever I try to come up with some arguments, I always feel like I am just arguing for the same attitudes in both countries, rather than the factor structure.
My question is: How do you make a hypothesis for this where you argue for the same/a different factor structure without arguing for the same/different attitudes?
Thank you in advance! :)
I came across a commentary titled, 'Tests of Statistical Significance – their (ab)use in the social sciences' and it made me reflect on the validity of using my sample for statistical testing. I have a sample of 24 banks and they were not randomly selected. They were the top 50 banks ranked by the Banker and I narrowed down the sample to 24 because only those banks were usable for my study. I wanted to test the association between these banks using a McNemar's test but any result I obtain- I obtained insignificant results - would be meaningless, right? Because they are not a random selection. I did not want to make a generalisation, but I wanted to know if I could still comment on the insignificance of their association?
Hello. We understand that a volcano plot is a graphical representation of differential values (proteins or genes), and it requires two parameters: fold change and p-value. However, for IP-MS (immunoprecipitation-mass spectrometry) data, there are many proteins identified in the IP (immunoprecipitation group) with their intensity, but these proteins are not detected in the IgG (control group)(the data is blank). This means that we cannot calculate the p-value and fold change for these "present(IP) --- absent(IgG)" proteins, and therefore, we cannot plot them on a volcano plot. However, in many articles, we see that these proteins are successfully plotted on a volcano plot. How did they accomplish this? Are there any data fitting methods available to assist in drawing? need imputation? but is it reflect the real interaction degree?
We measured three aspects (i.e. variables) of self-regulation. We have 2 groups and our sample size is ~30 in each group. We anticipate that three variables will each contribute unique variance to a self-regulation composite. How do we compare if there are group differences in the structure/weighting of the composite? What analysis should be conducted?
I have a paper that proposed a hypothesis test that is heavily based on existing tests (so it is pretty much a procedure built on existing statistical tests). It was rejected by a few journals claiming that it was not innovative, although I demonstrated that it outperforms some commonly used tests.
Are there any journals that take this sort of papers?
I want to ask about the usage of parametrical and non-parametrical tests if we have an enormous sample size.
Let me describe a case for discussion:
- I have two groups of samples of a continuous variable (let's say: Pulse Pressure, so the difference between systolic and diastolic pressure at a given time), let's say from a) healthy individuals (50 subjects) and b) patients with hypertension (also 50 subjects).
- there are approx. 1000 samples of the measured variable from each subject; thus, we have 50*1000 = 50000 samples for group a) and the same for group b).
My null hypothesis is: that there is no difference in distributions of the measured variable between analysed groups.
I calculated two different approaches, providing me with a p-value:
Option A:
- I took all samples from group a) and b) (so, 50000 samples vs 50000 samples),
- I checked the normality in both groups using the Shapiro-Wilk test; both distributions were not normal
- I used the Mann-Whitney test and found significant differences between distributions (p<0.001), although the median value in group a) was 43.0 (Q1-Q3: 33.0-53.0) and in group b) 41.0 (Q1-Q3: 34.0-53.0).
Option B:
- I averaged the variable's values over all participants (so, 50 samples in group a) and 50 samples in group b))
- I checked the normality in both groups using the Shapiro-Wilk test; both distributions were normal,
- I used t Student test and obtained p-value: 0.914 and median values 43.1 (Q1-Q3: 33.3-54.1) in group a) and 41.8 (Q1-Q3: 35.3-53.1) in group b).
My intuition is that I should use option B and average the signal before the testing. Otherwise, I reject the null hypothesis, having a very small difference in median values (and large Q1-Q3), which is quite impractical (I mean, visually, the box plots look very similar, and they overlap each other).
What is your opinion about these two options? Are both correct but should be used depending on the hypothesis?

Neurons were treated with four different types of drugs, and then a full transcriptome was produced. I am interested in looking at the effects of these drugs on two specific pathways, each with around 20 genes. Would it be appropriate for me to just set up a simple comparative test (like a t-test) and run it for each gene? Or should I still use a differential gene expression package like DESeq2, even though only a few genes are going to be analysed? The aim of my experiment is a very targeted analysis, with the hopes that I may be able to uncover interesting relationships by cutting out the noise (i.e., the rest of the genes that are not of interest).
During writing a review, usually published articles are collected from the popular data source like PubMed, google scholar, Scopus etc.
My questions are
1. how we can confirm that all the articles that are published in a certain period (e.g.,2000 to 2020) are collected and considered in the sorting process(excluding and including criteria)?
2. When the articles are not in open access, then how can we minimize the challenges to understand the data for the metanalysis?