Thank you for watching. You can find more statistics teaching videos here: ruclips.net/p/PLI7VLEjUJidCC7w6zpnhmYYj0XYuBy-Lt Please note that I do reply to questions/comments, but all previous responses have been lost in transitioning the channel over to my new email address.
the part that confused me was at 13:13 where you go from effect size .5 to effect size .8.....and the sample size decreased. Shouldn't it be the other way around? Where a larger effect size requires a larger sample size? thank you! :)
If I'm comparing TWO groups.... one standard of care (control) and the other intervention... looking at the same outcomes in both groups, would it still be diff between 2 independent means? It isn't matched pairs because it's still technically 2 different groups, right?
Sounds like you have a pre and post measure in each of 2 groups? So would want to compare the effects of the two interventions while controlling for baseline differences (perhaps - I'm not an expert in this).
Thank you very much for such a great clarification, just Can you let me know how to determine the sample size based on the previous study? I have two group cases and a control
Hi Karez. If you are able to use previous literature to determine the effect size that your study should be powered to detect, then you can follow the approach used in the video but with that effect size. You will also need to know the statistical test that you wish to use (i.e. the type of t-test, ANOVA, regression, etc.)
Hi! Thank you very much for your video! For my homework I have to calculate the sample size for my intervention study (I'm measuring t1 and t2) and I have one control group. So I guess that means I have to use the first way you described? Two independent groups? Or do I have to use another one?
Hi, Stuart! Really grateful for your video! My professor is asking us to familiarize ourselves with G*Power and wanted to confirm if what I am doing is correct? I have a population of 15,994 students and the research involves a two-tailed paired t-test. Since i'm interested in getting whether there is a significant change in the mean difference of students' scores before and after a certain phenomenon, I used the option "means: difference between two independent means (two groups)" under statistical tests. Additionally, I put 0.5 as the effect size, 0.05 alpha error, and 0.8 power in getting the sample size. Im really not sure if the values I used (especially in the effect size) is appropriate for the study considering that I have approximately 16,000 in the population and only 64 sample per group is required.
@@biomechstu thanks for pointing out that I should use dependent groups (matched pairs) instead of two independent groups! thank you also for linking the reference! Im able to better grasp the topic now :)) thank you so much!! 💓
Hello Stuart, I'm happy to watch your video. I'm carrying out a study of challenges with digital health solutions in therapy. My samples are patients. I intend to use regression analysis. How do I calculate the sample size please?
Hi Chinelo. You need to decide: the desired type I error rate (often 0.05); the effect size that you want to set the study up to detect (e.g. the smallest effect that would be clinically meaningful, or an expected effect based on previous studies); and the desired power or probability of detecting an effect at least as large as that one if it really exists (typically 0.80 or higher). These 3 values can then be used as inputs in G*Power as shown in the video, after selecting the regression analysis.
Hello Stuart. How about a 3 groups non directional test. We have 25 participants in each group, total of 75 and we would like to run the power analysis. Thanks.
Hi, thanks for your video. I was in trouble with my sample size calculation and I guess I found the answer in this video. I have just one question if I'm using the Chi-square test to check the correlation between two variables which test in the software I will choose? I couldn't find a chi-square in this calculator. Thanks very so much
Hi Amina. As a first (maybe) minor comment, chi-square isn't exactly a correlation test (unlike Pearson or Spearman's rank) but rather tests for independence of the two variables. But regardless, I don't have access to my computer at the moment but a Google search suggests a chi-square test is an option in G*Power. Maybe try some different menu options in case it's not under correlation.
Hi Stuart. We have recruited approx 270 participants which have been assessed on a specific questionnaire where we have divided the group into two (hence, now there are two 'independent' groups). I run the same G*power process as you have shown and found that we only need two groups of 64 each. When I change the allocation ratio to 0.7/0.8, I get Group 1: 78, and Group 2: 54 subjects. This means that we have already recruited sufficient number of participants? Thanks.
Great sharing, Tq dr. I just want to ask if I have two groups that I will match accordingly (they are not equal). One has 53 participants and the other is 51 participants. How to use the prior G*power?
Hello Stuart! Thank you for the video! I have a question about how I would calculate sample size/g-power for my study. I have on IV of quality of life and then DVs of (Satisfaction) via work, family, ADLs, and social perception from others . I will have one sample group. How would I run a g-power analysis on this? Thank you!
@@Dominique129 In that case, you can select the required ANOVA effect under 'Tests' and then 'Means' at the top of the window. You will need to know a few values such as your alpha (often 0.05), required power (often 80%), the number of factors/groups, and the size of effect that you want your study to be powered to detect. For example, that could be the smallest effect that you'd consider to be clinically or practically meaningful. Or it could be based on previous literature. Each of these values should ideally be justified but there are separate literature providing guidelines and recommendations for each if you wanted to go beyond default values, etc.
Thanks for the video. Could you please let me know how you can find out the statistical power, if you have recruited a set number of participants already? So for example want to know if 50 people you have recruited is enough for correlation and ANOVA
Hi Mario. Unless I've misunderstood, you're fine to follow the same guidance. i.e. figure out what statistical test is required, what your smallest effect size of interest is, your required alpha anf statistical power, and then follow the same steps. How you decide upon each of those parameters for the specific study is obviously the tricky part. Likewise, you may go for a completely different statistical and/or sample size justification approach.
Hi, I want to calculate sample size for developing an reference value ( normal value ) of endurance capacity using gpower. Which test should i apply to calculate sample size for this type of study. Plz do guide me
I'm sorry, I'm suffering cause I'm struggling with stadistics, I don't like investigation but I'm working on my thesis, could you help me please? I need to compare three groups, how the hell I do that?
I wanna calculate the sample size in ANOVA one way but i wanna calculate it with means and size data values. what must i suppose to be input in the table of means and size if i didn't have the data? from where i can get the data?
cThank you for this video. It is very helpful. I am doing a study in which I am using PLS (using SmartPLS software) analysis. My research model has 3 independent variables, 2 mediating variables and 1 dependent variable. Can I use G*power in this case to do power analysis? My sample size is 160. I need to show that power is not an issue with this sample size.
@@biomechstu thank you. What do you think about the method used in this paper- Aguirre-Urreta, M., & Rönkkö, M. (2015). Sample Size Determination and Statistical Power Analysis in PLS Using R: An Annotated Tutorial. Communications of the Association for Information Systems, 36, 3. doi:10.17705/1CAIS.03603
I'm not sure exactly what you mean, but this approach is purely for a sample size justification based on statistical power, not based on any alternative approach around generalisability or representative sampling, etc.
Hi Stuart, I am new to G Power. I want to conduct a survey of teachers. I know that the total population of primary teachers in my country is 28, 474. How do I calculate sample size based on this number to ensure I can generalise my results? I assumed that I would have to input this number somewhere. Thankyou
@@biomechstu Hi Stuart, I also have the same question. The only information I have is the population of civil servants in my study area which is about 17,945. Can you make a demo video how to calculate sample size based on this number?
Could you please tell estimated sample size assuming allele frequency 0.1 (10%), additive model, effect size is 2, and disease prevalence 7% with 0.05 % signig=ficance level
It looks like you have an effect size and an alpha level in mind. Just need a desired power and the statistical test that you plan to use. Then the approach in the video can be followed, if that is the desired approach for justifying sample size. Alternatively, if not a standard test then it may require simulation of data sets in R or similar, rather than G*Power
Hi, thx for the video! It's one thing I don't understand though. How is it possible to set both power and alpha since they are dependent of each other? decreasing alpha is increasing beta, leadning to a decrease in power? What am I missing here?
Hi. Power, alpha, effect size, and sample size are all inter-related. You can specify any three and determine the fourth. In this case, you are specifying alpha, power, and the effect size of interest and then determining the sample size that achieves these.
All 4 of power, alpha, effect size, and sample size affect each of the other three. This is why we have to pre-specify three of them and calculate the fourth. In your example, for any given effect size and sample size, changing the alpha will result in a change in power. Or for a given alpha, power, and effect size, you can work out the required sample size to achieve that.
Thank you, Stuart. How to choose effective size F (0.10 or 0.25 or 0.5) for One way ANOVA (one independent variable) in G power analysis. Please explain
How can I get a small portion of respondents from a specific amount of population? I have 351 for total population. I just want to get a portion of that population. How do I do that?
Hi. What statistical test would this be for? If your sample size justification is based onstatistical power, then out could follow the steps in the video. If the justification is based on generalisability or something else then it may be separate
If it's just for a statistical justification based on alpha, effect size, and statistical power, then you can follow the steps regardless of total population. If it is for some other reason then this may not be the required software/approach
What statistical test do you plan to use in your study? Once you know that, you just need to choose a desired alpha (e.g. 0.05), power (e.g. 80%), and minimum effect size of interest.
Seems like you erroneously state that the effect size box is for inputting the desired effect size but it is actually where you input the effect size you think your intervention or between-condition effect may be. This is why the sample size actually decrease when you increased the value in the effect size box. This is obviously a pretty important issue and the video should be edited to correct the issue.
Hi. This is the effect size you are powering your study to be able to detect if it exists. In some cases, this may be the exact size of effect that you expect but in most cases it won't be. It is often the smallest effect size that would be considered clinically or practically meaningful. For example, I may want to power the study adequately to detect effects of 0.2 or greater if they exist (for some a priori stated reason). This is regardless of what I actually think the effect size might be. As another example, if I anticipate an effect of 0.5 and set up the study to have 80% power to detect an effect size of 0.5 or greater, then my study would be underpowered to detect effects of 0.40 or 0.45, even if these effects are still large enough to be of clinical or practical significance. There are many ways of choosing an effect size such as (probably in order of preference): smallest effect size of interest; smallest effect size arbitrarily considered 'small', 'moderate', 'large', etc.; a central effect size estimate previously reported in the literature (but be aware this is only the central estimate and is also dependent on many factors in that study); a pilot test (but be aware of limitations in sample size of pilot work and the effect that can have on certainty of effect size estimates). I thin this is a long way of saying sometimes you may input the effect size you expect and other times you may input a different effect size for another valid a priori stated reason. The main thing is to justify the values chosen (including the alpha, beta, and effect size).
Thank you for watching. You can find more statistics teaching videos here: ruclips.net/p/PLI7VLEjUJidCC7w6zpnhmYYj0XYuBy-Lt Please note that I do reply to questions/comments, but all previous responses have been lost in transitioning the channel over to my new email address.
HELLO STUART, PLEASE WHICH EFFECT SIZE IS BEST FOR AN EXPERIEMENTAL STUDY?
Bless your soul!!! Simple, easy to understand, and to the point! Thank you so much!
Thank you! I really appreciate that.
Thank you so much for this. It really helps me. You go slowly through it explaining it and showing it visually which I need.
Thank you for the kind comments.
Simple and Informative, great video, very helpful. Keep on the good work!
Hi, How we can calculate participant size when population size is unknown. I want to use chi-square test. I need your guidance. Thank you
the part that confused me was at 13:13 where you go from effect size .5 to effect size .8.....and the sample size decreased. Shouldn't it be the other way around? Where a larger effect size requires a larger sample size? thank you! :)
Great stuff ! Maximum thanks
Great video... straight to the point!
Thank you!
If I'm comparing TWO groups.... one standard of care (control) and the other intervention... looking at the same outcomes in both groups, would it still be diff between 2 independent means? It isn't matched pairs because it's still technically 2 different groups, right?
Sounds like you have a pre and post measure in each of 2 groups? So would want to compare the effects of the two interventions while controlling for baseline differences (perhaps - I'm not an expert in this).
Thank you very much for such a great clarification, just Can you let me know how to determine the sample size based on the previous study? I have two group cases and a control
Hi Karez. If you are able to use previous literature to determine the effect size that your study should be powered to detect, then you can follow the approach used in the video but with that effect size. You will also need to know the statistical test that you wish to use (i.e. the type of t-test, ANOVA, regression, etc.)
Hi! Thank you very much for your video! For my homework I have to calculate the sample size for my intervention study (I'm measuring t1 and t2) and I have one control group. So I guess that means I have to use the first way you described? Two independent groups? Or do I have to use another one?
So Helpful! Thanks a lot!
Thank you very much; it is something I have been looking for and to informative
Great thank you very much
Could you please tell me sample size determination for animal study using g power?
did you get the response? i need guidance on how to determine a sufficient fish sample size to determine the length distribution of the population
Super helpful many thanks:)
Hi, Stuart! Really grateful for your video! My professor is asking us to familiarize ourselves with G*Power and wanted to confirm if what I am doing is correct? I have a population of 15,994 students and the research involves a two-tailed paired t-test. Since i'm interested in getting whether there is a significant change in the mean difference of students' scores before and after a certain phenomenon, I used the option "means: difference between two independent means (two groups)" under statistical tests. Additionally, I put 0.5 as the effect size, 0.05 alpha error, and 0.8 power in getting the sample size. Im really not sure if the values I used (especially in the effect size) is appropriate for the study considering that I have approximately 16,000 in the population and only 64 sample per group is required.
@@biomechstu thanks for pointing out that I should use dependent groups (matched pairs) instead of two independent groups! thank you also for linking the reference! Im able to better grasp the topic now :)) thank you so much!! 💓
Hello Stuart, I'm happy to watch your video. I'm carrying out a study of challenges with digital health solutions in therapy. My samples are patients. I intend to use regression analysis. How do I calculate the sample size please?
Hi Chinelo. You need to decide: the desired type I error rate (often 0.05); the effect size that you want to set the study up to detect (e.g. the smallest effect that would be clinically meaningful, or an expected effect based on previous studies); and the desired power or probability of detecting an effect at least as large as that one if it really exists (typically 0.80 or higher). These 3 values can then be used as inputs in G*Power as shown in the video, after selecting the regression analysis.
Thank you so much
Hello Stuart. How about a 3 groups non directional test. We have 25 participants in each group, total of 75 and we would like to run the power analysis. Thanks.
Great Stuart . Thanks !!
thanks, so simple :)
Ahhh, the G-spot!!
Hi Stuart, I wonder if G*Power works for two-/multi- stage cluster sampling technique as well?
Hi, thanks for your video. I was in trouble with my sample size calculation and I guess I found the answer in this video. I have just one question if I'm using the Chi-square test to check the correlation between two variables which test in the software I will choose? I couldn't find a chi-square in this calculator.
Thanks very so much
Hi Amina. As a first (maybe) minor comment, chi-square isn't exactly a correlation test (unlike Pearson or Spearman's rank) but rather tests for independence of the two variables. But regardless, I don't have access to my computer at the moment but a Google search suggests a chi-square test is an option in G*Power. Maybe try some different menu options in case it's not under correlation.
How about calculating power within vignette experiments?
Hi Stuart. We have recruited approx 270 participants which have been assessed on a specific questionnaire where we have divided the group into two (hence, now there are two 'independent' groups). I run the same G*power process as you have shown and found that we only need two groups of 64 each. When I change the allocation ratio to 0.7/0.8, I get Group 1: 78, and Group 2: 54 subjects. This means that we have already recruited sufficient number of participants? Thanks.
Hi how to calculate sample size for a cross sectional study using g power
Great sharing, Tq dr.
I just want to ask if I have two groups that I will match accordingly (they are not equal). One has 53 participants and the other is 51 participants.
How to use the prior G*power?
Hello Stuart! Thank you for the video! I have a question about how I would calculate sample size/g-power for my study. I have on IV of quality of life and then DVs of (Satisfaction) via work, family, ADLs, and social perception from others . I will have one sample group. How would I run a g-power analysis on this? Thank you!
Hi Dominique. Do you know what statistical test you intend to use for your hypothesis testing?
@@biomechstu An ANOVA I believe. Thank you for responding!
@@Dominique129 In that case, you can select the required ANOVA effect under 'Tests' and then 'Means' at the top of the window. You will need to know a few values such as your alpha (often 0.05), required power (often 80%), the number of factors/groups, and the size of effect that you want your study to be powered to detect. For example, that could be the smallest effect that you'd consider to be clinically or practically meaningful. Or it could be based on previous literature. Each of these values should ideally be justified but there are separate literature providing guidelines and recommendations for each if you wanted to go beyond default values, etc.
@@biomechstu Thank you so much. I sent a follow-up email just for further clarification.
How to calculate affect size with the help of sample size .. how to add values in sample size?
What could an interpretation be about the blue dotted distribution and the non-centrality parameter?
Thanks for the video. Could you please let me know how you can find out the statistical power, if you have recruited a set number of participants already? So for example want to know if 50 people you have recruited is enough for correlation and ANOVA
@@biomechstu Perfect! Thanks.
Hi, can you show how to calculate a sample size for an epidemiological study? Thank you!
Hi Mario. Unless I've misunderstood, you're fine to follow the same guidance. i.e. figure out what statistical test is required, what your smallest effect size of interest is, your required alpha anf statistical power, and then follow the same steps. How you decide upon each of those parameters for the specific study is obviously the tricky part. Likewise, you may go for a completely different statistical and/or sample size justification approach.
Hi,
I want to calculate sample size for developing an reference value ( normal value ) of endurance capacity using gpower. Which test should i apply to calculate sample size for this type of study. Plz do guide me
I'm sorry, I'm suffering cause I'm struggling with stadistics, I don't like investigation but I'm working on my thesis, could you help me please? I need to compare three groups, how the hell I do that?
Really helpful video. Thank you.
I wanna calculate the sample size in ANOVA one way but i wanna calculate it with means and size data values. what must i suppose to be input in the table of means and size if i didn't have the data? from where i can get the data?
cThank you for this video. It is very helpful.
I am doing a study in which I am using PLS (using SmartPLS software) analysis. My research model has 3 independent variables, 2 mediating variables and 1 dependent variable. Can I use G*power in this case to do power analysis? My sample size is 160. I need to show that power is not an issue with this sample size.
@@biomechstu thank you. What do you think about the method used in this paper- Aguirre-Urreta, M., & Rönkkö, M. (2015). Sample Size Determination and Statistical Power Analysis in PLS Using R: An Annotated Tutorial. Communications of the Association for Information Systems, 36, 3. doi:10.17705/1CAIS.03603
Hi Stuart, can I arrange for an asap immediate private consultation on how to calculate my sample size from you please?
How do you determine the G Power for Spearman Correlation rho? One group
Hi my research population is 32.7 million. I have 4 independent variable and 1 dependent variable. How can I determine my sample size using Gpower?
My research population at 43.5 million. How can i justify my sample size using GPower?
Will number of variables have any effect on the sample size?
Does the population matter? Where do you input the population?
I'm not sure exactly what you mean, but this approach is purely for a sample size justification based on statistical power, not based on any alternative approach around generalisability or representative sampling, etc.
hi how to calculate sample size for intervention based study ?
Hi Stuart, I am new to G Power. I want to conduct a survey of teachers. I know that the total population of primary teachers in my country is 28, 474. How do I calculate sample size based on this number to ensure I can generalise my results? I assumed that I would have to input this number somewhere. Thankyou
@@biomechstu Hi Stuart, I also have the same question. The only information I have is the population of civil servants in my study area which is about 17,945. Can you make a demo video how to calculate sample size based on this number?
Could you please tell estimated sample size assuming allele frequency 0.1 (10%), additive model, effect size is 2, and disease prevalence 7% with 0.05 % signig=ficance level
It looks like you have an effect size and an alpha level in mind. Just need a desired power and the statistical test that you plan to use. Then the approach in the video can be followed, if that is the desired approach for justifying sample size. Alternatively, if not a standard test then it may require simulation of data sets in R or similar, rather than G*Power
Hi, thx for the video! It's one thing I don't understand though. How is it possible to set both power and alpha since they are dependent of each other? decreasing alpha is increasing beta, leadning to a decrease in power? What am I missing here?
Hi. Power, alpha, effect size, and sample size are all inter-related. You can specify any three and determine the fourth. In this case, you are specifying alpha, power, and the effect size of interest and then determining the sample size that achieves these.
@@biomechstu ok, but I'm still confused though:/ so alpha does not directly affect beta and power?
All 4 of power, alpha, effect size, and sample size affect each of the other three. This is why we have to pre-specify three of them and calculate the fourth. In your example, for any given effect size and sample size, changing the alpha will result in a change in power. Or for a given alpha, power, and effect size, you can work out the required sample size to achieve that.
Thank you, Stuart. How to choose effective size F (0.10 or 0.25 or 0.5) for One way ANOVA (one independent variable) in G power analysis. Please explain
If I take more the number of calculated g power, is it fine
how do you know a priori the effect size? I am planning to do some experiments and for that I need sample size. Pls comment.
How do I calculate the sample size for a four arm RCT.
Hi Caroline. Do you know what statistical test you plan to use within your analysis?
can u help with ANCOVA calculation?
Hi. Sorry for the slow reply. Do you have a specific question, or one of the values that you're unsure about?
is this applicable for mediation analysis?
How can I get a small portion of respondents from a specific amount of population? I have 351 for total population. I just want to get a portion of that population. How do I do that?
Hi. What statistical test would this be for? If your sample size justification is based onstatistical power, then out could follow the steps in the video. If the justification is based on generalisability or something else then it may be separate
@@biomechstu I'll be using it to get a small percentage of a total population for research
If it's just for a statistical justification based on alpha, effect size, and statistical power, then you can follow the steps regardless of total population. If it is for some other reason then this may not be the required software/approach
Hi I could use your help determining a sample size for a research study
What statistical test do you plan to use in your study? Once you know that, you just need to choose a desired alpha (e.g. 0.05), power (e.g. 80%), and minimum effect size of interest.
Seems like you erroneously state that the effect size box is for inputting the desired effect size but it is actually where you input the effect size you think your intervention or between-condition effect may be. This is why the sample size actually decrease when you increased the value in the effect size box. This is obviously a pretty important issue and the video should be edited to correct the issue.
Hi. This is the effect size you are powering your study to be able to detect if it exists. In some cases, this may be the exact size of effect that you expect but in most cases it won't be. It is often the smallest effect size that would be considered clinically or practically meaningful. For example, I may want to power the study adequately to detect effects of 0.2 or greater if they exist (for some a priori stated reason). This is regardless of what I actually think the effect size might be. As another example, if I anticipate an effect of 0.5 and set up the study to have 80% power to detect an effect size of 0.5 or greater, then my study would be underpowered to detect effects of 0.40 or 0.45, even if these effects are still large enough to be of clinical or practical significance. There are many ways of choosing an effect size such as (probably in order of preference): smallest effect size of interest; smallest effect size arbitrarily considered 'small', 'moderate', 'large', etc.; a central effect size estimate previously reported in the literature (but be aware this is only the central estimate and is also dependent on many factors in that study); a pilot test (but be aware of limitations in sample size of pilot work and the effect that can have on certainty of effect size estimates). I thin this is a long way of saying sometimes you may input the effect size you expect and other times you may input a different effect size for another valid a priori stated reason. The main thing is to justify the values chosen (including the alpha, beta, and effect size).