Ask Power - Recent questions

An error occured. Please check your input or contact us with the code 373368653a8043b6b945e45f7da0fb1c.

Fri, 29 May 2026 02:01:02 +0000

An error occured. Please check your input or contact us with the code c5e82b37570d75c49c6a748c12ddb959.

Thu, 21 May 2026 05:34:04 +0000

what is An error occured. code fc9b0a8833a17c4efc560d684353e3eb.

Mon, 18 May 2026 17:16:16 +0000

Error for output of skewness and kurtosis calculation

Wed, 13 May 2026 08:59:05 +0000

Output of skewness and kurtosis calculation

An error occured. Please check your input or contact us with the code 3f13029984c9c99e5f87ce81fc89bc8e.

clarification of input for moderated mediation model 7

Thu, 06 Mar 2025 07:49:13 +0000

Dear WebPower team,

Thanks for this great tool!

I would like to ask clarification on the following inputs for model 7 moderated mediation.

1. Are the regression coefficients supposed to be standardized betas?

2. How to decide on the moderator value?

More generally, it is not always clear to me in WebPower which measures of effect size I should input.

Thanks very much in advance.

Best wishes,

Lukas Van Oudenhove

No such analysis!

Wed, 25 Sep 2024 20:45:42 +0000

i used webpower to check Mardia multivariate kurtosis analysis before one year and a half, today i opened the results link it gave me "No such analysis". does this mean that powerweb erased the data? can i retrieve it?

No calculation for mediation analysis models (moderate mediation)

Mon, 26 Aug 2024 17:59:18 +0000

Hello! I am experiencing problems in the analysis of moderated mediation, both in ‘Model 7 ConditionalProcesses UsingJointSignificance’ and ‘Model 8 Conditional Processes Using Joint Significance’. Specifically, after entering the parameters, clicking on ‘calculate’ does not seem to happen, as if the command is not received. I have also tried changing browsers but the situation is the same. Thank you very much if you can help me!

State when N refers to N per group or total N for each analysis

Mon, 03 Jun 2024 08:12:41 +0000

It would be very helpful if the output screen could clearly state if N refers to N per group or total N for all groups combined. I haven't tried many tests, but my impression is that this not always consistent across tests - and even if it was, it would still be helpful.

Many thanks for the WebPower package - it is awesome

Model 15 Power calculation no ouput

Thu, 23 May 2024 12:38:39 +0000

Hello, first of all I'm very thankful for your Power Analysis tool for moderated Mediations. For Model 14, everything seems to work just fine, there's an output after like a minute. But for Model 15 , there seems to be no output for like half an hour (and the caluclate button animation is resetting as there might should be an output). Is the compute for this model that heavy, or could this be a bug with Model 15? (or I might do something wrong)

Please, Am I doing right?

Tue, 21 May 2024 20:30:45 +0000

Hi, I'm Young, This is my model specification for determining sample size and power. But, it didn't run this way, Did I input the wrong way?

Thank you.

Trouble uploading data to effect size calculator

Wed, 17 Apr 2024 15:21:55 +0000

Hi! I have tried uploading some data to WebPower to use the effect size calculator (https://webpower.psychstat.org/models/means05/effectsize.php), but have not been able to get my data uploaded (see section 4). The section for uploading doesn't specify which data format to use, but when I've uploaded repeated measures ANOVA data (3 columns, representing conditions, and 18 rows, representing participant ids) in .csv, .mat, and .txt format, none have worked.

The data appears to upload, but nothing happens when I click "Calculate". I've tried this function on both Safari and Chrome, in case there was a problem with the upload.

Thanks for any help on this front!

Best,

Anna

How to cite and provide reference for Univariate and multivariate skewness and kurtosis calculation?

Wed, 03 Jan 2024 15:25:05 +0000

Hello,

I want to cite the tool "Univariate and multivariate skewness and kurtosis calculation" in my manuscript. How to cite it and what is the reference for this tool?

Thanks

WebPower removed from CRAN?

Wed, 20 Jul 2022 15:13:31 +0000

Hello,

I noticed WebPower was removed from CRAN on 7/19/2022. Is there a plan to add it to CRAN again?

Thanks

An encountered error when checking the results of the diagram based power analysis

Sat, 28 May 2022 12:56:33 +0000

I report the following error. Please take care of it.

ERROR: An error has been found in your power analysis. Please send us the URL to your analysis day=20220528&base=ef6323e91eb8115984788fe8828ad4cb so that we can fix it. Thanks.

calculate sample size

Sat, 23 Apr 2022 00:04:14 +0000

Poisson power analysis

Mon, 07 Feb 2022 21:23:33 +0000

Hi,

We're conducting a power analysis using wp.poisson to get the relative risk for different levels of a normally distributed predict (mean = 0, sd = 1). Our dependent variable is binary, but it was suggested that we use a poisson rather than logistic since the prevalence of the behavior we're observing is over 10% (22-40% has been observed in previous studies). The observed intercept in simulated data is -1.91, based on the 22% occurrence, so the IRR is 0.148. I used the relative risk for a small, medium and large effect as 1.22, 1.86, and 3.00, as suggested by Olivier et al (2017). I get an error when I try to run the medium and large effects of "non-finite function value." Here's my code:

medium <- wp.poisson(n = seq(40, 100, 10), exp0 = 0.15, exp1 = 1.86, alpha = 0.05, power = NULL, family = "normal", parameter = c(0, 1))

medium

Any suggestions for trouble shooting?

Power Curve for multigroup mediation models, non normal data and a continious moderator

Sun, 23 Jan 2022 12:48:29 +0000

- Is it possible to calculate a power curve for multigroup mediation models in R?

- Is it possible to take account for non normal data in R when x is dichotomous?

- Is there another way to calculate power for an moderated mediation model when the moderator is continious than to use +/- 1SD of the moderator (w)? I could calculate Johnson-Neyman intervals for the x-w-y realationship and put the group split at the first significant value of the moderator for example.

Thank you and kind regards,

Alex

How can I fix the negative beta value in moderation analysis?

Thu, 06 Jan 2022 18:56:49 +0000

I am examining the moderating role of transformational leadership style (TLS) in the relationship between knowledge sharing (KS) and innovative work behavior (IWB). For this purpose, I have adopted a 25 items questioner, and I have received 250 responses.

in the SPSS 23 analysis, I run the "linear regression" to test the impact of moderating variables in the relationship between DV and IV. as a result in the "Coefficient" table shows a negative value of "-1.303" for "Beta Standard Coefficient" while the P-value is .000 "significant". (Photo attached in the below)

So, my question is: How can I get a positive value for "beta"? also which one of the variables needs to be corrected?

Thank you.

Problems with wp.crt2arm

Sat, 11 Sep 2021 20:12:47 +0000

Hello, I cannot calculate the number of clusters needed in a Cluster randomized trial, using wp.crt2arm in R. I don't know if it's a software problem. The message is as follows:

> wp.crt2arm(f=0.6,n=30,J=NULL, icc=.2,power=.8)

Error in nuniroot(function(J) eval(p.body) - power, ifelse(is.null(interval),: Please provide an interval with two values such as c(0,1).

A question on the syntax for power analysis using the "wp.mc.sem.basic" function

Mon, 24 May 2021 08:03:16 +0000

Please let me illustrate my question with a concrete example of power analysis for the following model (some irrelevant parts have been omitted to save space):

library (WebPower)

model.2f <- '

F1 =~ a*x1 + start (.55)*x1 + start (.7)*x2 + start (.7)*x3

F2 =~ b*x4 + start (.6)*x4 + start (.6)*x5 + start (.5)*x6

F1 ~~ start (1)*F1

F2 ~~ start (1)*F2

F1 ~~ start (.3)*F2

x1 ~~ start(.6)*x1

…

power.2f <- wp.mc.sem.basic(model = model.2f, indirect = NULL, nobs = 100, nrep = 5000)

summary (power.2f)

In the outputs, the power analysis results (MSE, SD, Power, Coverage) for the indicator variables, x1 and x4, of the two respective factors in the model are not available, though I have set both factors’ variances as 1 and the loadings of x1 and x4 as freely estimated. By “not available” I mean their corresponding output rows are like:

True Estimate MSE SD Power Coverage

F1 =~ x1 (a) 0.550 0.550 0.000 0.000 NaN 0.000

…

F2 =~ x4 (b) 0.600 0.600 0.000 0.000 NaN 0.000

Grateful if you could advise how the power analysis results can be displayed for ALL the indicator variables (x1 – x6 in this case) of a measurement model when the latent variables they indicate have been set to have unit variances.

Power analysis for data-model fit using a Monte-Carlo approach

Mon, 24 May 2021 02:47:09 +0000

Grateful if your team could advise whether the WebPower or any other R package(s) may provide power analysis facilities using a Monte-Carlo approach for the fit indices (e.g., AIC, SRMR) of a given structural equation model. For example, Mplus provides Monte Carlo analysis outputs for the expected and observed distributions of the fit indices of a specified model. It will be great if you can point me to similar facilities/functionalities available via WebPower or other R packages that are free. (I understand the functions of WebPower like "wp.sem.chisq", "wp.sem.rmsea" provide power analysis for fit statistics/indices - I am specifically interested in the availability of the corresponding facilities using a Monte Carlo approach).

how to perform sensitivity power analysis for independent samples t-test using our effect size?

Sun, 23 May 2021 16:04:49 +0000

hi there,

i used one way between groups ANOVA for my data, 3 groups, the overall result was non significant, so I did not follow any planned contrasts ( 2 planned contrasts).

however, I was told that I may need to run 2 a priori analysis for indenpendent samples t test for two groups, two tailed , but because the data is already collected then I have to run sensitivity power based on the sample I was given. I have 2 effect size for both groups ( anger group 0.08, fear group, 0.10)

I uploaded my data file www.webpower.psychstat.org/models/means03/effectsize.php

regards

how to determinate ng and nm in wp.rmanova

Mon, 17 May 2021 23:56:36 +0000

Hi,

I am calculating the needed sample size for a new experiment using wp.rmanova function. The experiment is a fully crossed within-subject 2x2 design. I have two factors, and each factor has two levels. It is a signal detection type of experiment. For each condition, it has 180 trials. So, total 180 * 4 = 720 trials for each participant.

I am confused about what the inputs for ng and nm in this situation are. According to the manual, ng seems to be the number of between-subject factors, and nm seems to be the levels of the within-subject factors.

Thus, should the parameters be ng = 1 and nm = 4, ng = 1 and nm = 2, or considering each trial is one measurement, ng = 1 and nm = 720?

Thank you for your help!

Simple mediation of fixed effects for random intercept models

Tue, 25 Aug 2020 01:55:17 +0000

Is it possible to use the path analysis tool for mediation models with repeated measures? I am looking to power a simple mediation model, but one where the two regressions have random intercepts. In other words, there is an independent variable X, a mediator M that every participant responds to n times, and a dependent variable Y that every participant responds to n times. I know I can simply use OLS if I average over the n responses, but I was hoping to avoid that. Thank you in advance for your help.

Multinomial logistic regression - Power analysis

Fri, 03 Jul 2020 13:19:30 +0000

I am trying to estimate the sample size (power = 80; alpha = 0.05) required for a multinomial logistic regression.

The IV (x) is a dummy variable (0,1). The DV (y) is a nominal variable with 4 categories (0,1,2,3).

The hypothesis is that when x = 0, there would be an equal chance of observing any of the categories in y (i.e., .25 chance per category). And when x = 1, there will be a .60 chance of observing the specific category (0) in y.

The wp tutorial seems to speak only to binomial logistic regressions. Any guidance on how to determine a sample size will be much appreciated.

Power calculation for model with ordinal outcome variable

Fri, 01 May 2020 16:06:40 +0000

Hello,
I am conducting a power calculation for a Stan model using brm() (https://www.rdocumentation.org/packages/brms/versions/2.12.0/topics/brm) and wp.regression(). My model has an ordered factor as the outcome variable and therefore uses family ‘cumulative’. However, I am having trouble calculating the model's R^2 statistic because the function bayes_R2() does not accept input from a model with an ordinal outcome variable / cumulative family type. Is there another way to compute the R^2 statistic for this type of model?

Does wp.regression() alpha parameter specify a one-tailed or two-tailed test?

Fri, 01 May 2020 15:56:27 +0000

I am using the function wp.regression() and setting the alpha parameter to 0.05. However it's not clear whether it is computing a one-tailed or two-tailed test. Which test is default, and can I specify that one-tailed is required?

Effect size in RM ANOVA: f or d?

Thu, 23 Apr 2020 19:16:33 +0000

Hi,

I have been trying to determine the sample size for a repeated measures design with one repeated measures variable with two levels.

First, I tried with the t-test (https://webpower.psychstat.org/models/means01/), with power = 0.80, alpha = 0.05, two sided, type = paired, and a medium effect size of d = 0.50 (I assumed that the effect size was Cohen's d because of the outcome of the analysis and the man page of the R package). The result is n = 33.37, which is consistent with pwr package in R.

I also tried to replicate the analysis above with a repeated measures ANOVA (https://webpower.psychstat.org/models/means05/). My understanding is that for a within participants design with one variable with two levels, if parameters are the same the results (i.e., the expected n) should be the same as with the t-test.

For one group and two measures (that is, only one RM factor with two levels), power = 0.80, type of effect = within, and leaving sample size empty, for a medium effect of Cohen's f = 0.25 (equivalent to a d = 0.50) the outcome is that I would need 127.5 participants. Interestingly, if I enter f = 0.50 the outcome is n = 33.37, as with the t-test.

It also happens the other way round: in the t-test power analysis, if we introduce d = 0.25 the resulting n is 127.5. Thus, it seems that Cohen's d in the t-test and Cohen's f in the RM ANOVA are the same, when actually they are different (f = d / 2).

I wonder if I misunderstood the effect size of the RM ANOVA and, instead of f, it is a d, or there is any other difference between the t-test and the RM ANOVA in this particular case. Could you, please, confirm whether in the RM ANOVA we should enter Cohen's f or Cohen's d as effect size? If it's Cohen's f, why the discrepant results between the t-test and the RM ANOVA?

Thanks for your time.

How do i unsubscribe?

Fri, 13 Mar 2020 02:07:34 +0000

Hi there, I would like to unsubscribe

Thanks

post hoc power analysis

Tue, 18 Feb 2020 20:43:46 +0000

Hi

I tested the simple mediation model in the manual (page 287):

Initially, I defined the parameter estimates as follows:

a @ .39 b @ .39 and cp @ 0, but the program gave an error.

later

a? .39 b? .39 and cp? 0, but the program gave an error again.

I gathered data from a small group to determine the sample size for my study (n = 117). Firstly, I analyzed this data. I want to add the parameter estimates I obtained from this analysis as fixed values in the model, and I want to calculate power according to these values.

@ or ? instead of using *, the program runs, but the parameter estimates on the output page are not what I initially defined; Estimates close to 0.

What do you suggest in this case? I want to know if I should add to my sample by doing post hoc power analysis.

Thank you.

Zeynep

Please send us the URL to your analysis day=20200207&base=67d262748ad87d3963c776f660248cad so that we can fix it.

Fri, 07 Feb 2020 14:32:10 +0000

Hello,

I want to calculate power for the 65th model of Dr. Hayes. I used Hayes'symbols. I added following interactions to "Power parameter section":

a1b3: = a1 * b3

a1b4: = a1 * b4

a3b1: = a3 * b1

a3b3: = a3 * b3

a3b4: = a3 * b4

I also added covariance to among the independent variables in the model, described all paths and covariances as "? .10".

Webpower gives a calculation error. What could be the source of the problem?

Thank you.

Sample Size for binomial distribution

Tue, 25 Jun 2019 20:44:58 +0000

Scenario:

Sensitivity of HCV test is .995 (result is binary, positive/negative)

Power .9

Precision .01

Prevalence is .01

Alpha .05

How many people do I have to screen at the 95% and then at the 99% CI to establish the sensitivity of 99.5%?

Citing Cohen et al for calculation of effect size in 3 arm CRT

Sun, 28 Apr 2019 23:28:42 +0000

Hi, is the effect size calculation for 3-arm cluster randomized trial as outlined in "Practical Statistical Power Analysis using WebPower and R" based on Cohen d calculation for effect sizes? I ask because I need to outline that information in my writing, together with citing your work. Please advise.

Effect-size ANOVA equation

Thu, 04 Apr 2019 19:07:43 +0000

Dear

Firstly, I congratulate the initiative on facilitating the analysis of the data through the website.

I want to analyze the size of the effect through the link site https://webpower.psychstat.org/models/means03/effectsize.php

Could you offer information on the equation applied in the link?

Sincerely

Effect Size for 3-arm cluster randomized trial

Sun, 17 Mar 2019 08:35:08 +0000

Hi,

I am in the process of calculating the effect size for a 3-arm cluster randomized trial. I was successful in calculating it using your software. However, I would like to see the equations that enabled the calculation, so I can repeat it in SAS and references for the equations as I need it for a paper I am writing.

Thanks!

Representing interaction terms in montecarlo SEM power

Tue, 05 Mar 2019 16:37:41 +0000

Hello,

I've been rethinking my previous question and I think, I would be able to solve that one if I knew how to represent interaction terms between exogenous variables with lavaan based syntax for a sem montecarlo power estimation.

Is there anyway I can do it?

For example when dealing with moderated mediation?

Sample size for a 3-way moderation and moderated mediation

Tue, 05 Mar 2019 11:22:48 +0000

Good morning,

Could you please help me out in a setting a template (R syntax) to simulate power at various sample sizes for

mod3 <- '
# Model 3 PROCESS Hayes - 3way interaction
# XW, XZ, WZ, XWZ are products of predictors/moderators
Y ~ b1*X + b2*W + b3*Z + b4*XW + b5*XZ + b6*WZ + b7* XWZ
'
fit<-lavaan::sem(mod3, df, se = "boot")

X, W and Y are continouse
Z is a 2 level factor, coding experimental conditions (treatment vs control)

I fitted this model on pilot study data where a 3way interaction was hypothesized. Of course I had to manually compute product variables for interaction terms

Now I'd like to run a sample size simulation for an other study, where both Z and W will be experimental conditions (2x2 between subjects) and the effect size is expected to by small-medium.
I can't figure out, how to handle the fact that XZ, XW, ZW, XWZ are not distinct measured variables but products.

I'd appreciate your help in sorting this out.

The lab i'm working in, deals with (in Andrew Hayes PROCESS terminology) moderations (model1 and 3), simple mediation (model4 - that's easy), and moderated mediations (model 7 and model 14).

I'd love to sort out a workable simulation template for doing sample size calculations at (conservatively) low effect sizes.

Thank you

multilevel modeling power analysis

Sat, 29 Sep 2018 17:06:48 +0000

Hi, I have a two-level nested-design dataset for power analysis. Level 1 represent goals we have subjects listed; level 2 represent differences between individuals. So it is goal nested in person design. The tricky thing is that we do not have treatment/control group; all variables are continuous. Not sure if wecan use WebPower to calculate power.

accounting for clustered data

Mon, 30 Jul 2018 20:42:06 +0000

Hi! I'd like to do a power analysis for latent-variable path model (SEM). I will be using clustered data (assessment scores from about 260 children who were in clusters of about 6 children per classroom). I plan to use the option in Mplus to adjust standard errors for clustering, as I'd like to avoid multilevel modeling. Is there a way to account for this clustered data in webpower? Thank you!

multivariate skewness and kurtosis online calculation excel file cant be open

Thu, 17 May 2018 02:44:17 +0000

Hi,

I have used this webpower to test skewness and kurtosis using the online calculator before. However, since yesterday, my excel file uploaded, when click calculate, it shows 'can't open file'.

What is causing this? How can i measure my data normality?

Hope someone can assist me here. Thanks

How to perform a post hoc test after completed one-way ANOVA with binary data

Sat, 21 Apr 2018 15:41:14 +0000

After completed a one-way ANOVA with binary data from the study "Statistical Power Analysis for One-way ANOVA with Binary or Count Data" we would like to performe a post hoc test. Do you have any recommendations for post hoc tests that would be suitable?

Many thanks,

Felicia

Could you please guide me in calculating skewness and kurtosis (Mardia's multivariate coefficient): load data, etc

Sun, 27 Aug 2017 00:17:28 +0000

I have been struggling to upload my data to enable me calculate mulitivariate normality using Mardia's coefficient. Could you please guide or help me with instructions on how to do this?

Many thanks,

Swala

Where can I download the WebPower package for R?

Thu, 20 Apr 2017 21:04:34 +0000

I would like to download the WebPower package so that I can incorporate the power analysis code directly into my R scripts. Is this possible or does it only run on a server? If it is possible where I download the package?

A link is provided for a pdf of the power curve, even if the pdf is empty.

Thu, 05 May 2016 14:59:47 +0000

When getting used to using the software, I tried to ask for a power curve while only one sample size value was provided (laterI figured out that the user has to provide multiple sample sizes for the plot). A link for a power curve pdf shows up in the results, however it will not open because the pdf has no pages.

But why you equalize S with the standard error (S.E.)?

Mon, 14 Mar 2016 08:08:57 +0000

In the formulae of Sp, S_1^2 and S_2^2 it refers to standard deviation of each population, why you equalize to the standard error?.

Thanks for your time Johnny!!!

but in that case, where do I put the mean values of my samples?

Thu, 10 Mar 2016 14:47:48 +0000

Thanks Johnny!....but, If I leave the effect size blank, where do I put the values of my means and the standard deviation?.

Can I estimate the detectable differences with Web Power?

Thu, 10 Mar 2016 13:48:04 +0000

Hello, Im trying to perform a post-hoc analysis to estimate the detectable differences (percentage difference of detectable

treatment means relative to control means) for a = 0.05 and a power of 80%. Does Web Power do this?...I have been

trying but cant find the correct way. Want to estimate the detectable differences of a t-test (two tailed).

If anyone can help me it would be great!!

My data are:

sample 1) N= 16; mean= 122.4; S.E ± 8.38

Sample 2) N = 20; mean= 111.4; S.E. ± 6.35

Thanks in advance!

Matías

How to cite WebPower

Fri, 08 May 2015 01:26:02 +0000

If you would like to cite WebPower, please use the following:

Zhang, Z., & Yuan, K.-H. (2018). Practical Statistical Power Analysis Using Webpower and R (Eds). Granger, IN: ISDSA Press. [https://webpower.psychstat.org]

How to register to be a user?

Tue, 04 Nov 2014 12:16:08 +0000

How to start a new power analysis?

Tue, 04 Nov 2014 12:11:34 +0000

How to use WebPower to start a new power analysis?