[tips] Sample Size: How to Determine it?
I'm reading an interesting piece of research on anthropomorphism which essentially states after a natural disaster if we use the term mother nature when describing it, people will be less willing to contribute to relief efforts (Humanizing nature could help the perceiver to conceive natural events as imbued with intentionality and significance rather than considering them merely random and meaningless phenomena). They did two studies. Here's the issue/question: Study 1 was correlational and involved 96 students. The results were supportive at .001 Study 2 was an experiment (no need to go into the details) involving 56 students. The results were, in the authors words, tangentially supportive with p.06 I think the study was well conducted so I don't mean to slight the researchers. My guess is that if they used more subjects they probably would have reached p.05 - but would that have been an example of selective stopping? I assume it would be. So how exactly does a researcher determine beforehand - as we are suggesting they do - the number of subjects they ought to try to get for the study? I'm just not familiar with the process. Does one look at the effect sizes of previous related studies to determine if the effect is large or small and then make a decision? But let's say the effect is assumed to be small, so do you use 100 subjects? 500? How is this number determined? Appreciate the insight in this. Michael Michael A. Britt, Ph.D. mich...@thepsychfiles.com http://www.ThePsychFiles.com Twitter: @mbritt --- You are currently subscribed to tips as: arch...@jab.org. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5n=Tl=tipso=27372 or send a blank email to leave-27372-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
Re: [tips] Sample Size: How to Determine it?
There is software to determine this. One excellent and free app is G*Power. http://www.psycho.uni-duesseldorf.de/abteilungen/aap/gpower3/ I would use the correlational study to give me an estimate of effect size. As you describe, I would use that in the software to estimate my number of participants to attain the desired power. Practicality constraints on number of available participants usually limits things. I did such an estimate using G*Power a few weeks ago for a study we are planning. We will need to collect data over two semesters because the anticipated number of participants available from one semester's worth of students would only give us power of about .66, whereas two semester's worth would bump us up over .90. Paul On Aug 27, 2013, at 8:18 AM, Michael Britt wrote: I'm reading an interesting piece of research on anthropomorphism which essentially states after a natural disaster if we use the term mother nature when describing it, people will be less willing to contribute to relief efforts (Humanizing nature could help the perceiver to conceive natural events as imbued with intentionality and significance rather than considering them merely random and meaningless phenomena). They did two studies. Here's the issue/question: * Study 1 was correlational and involved 96 students. The results were supportive at .001 * Study 2 was an experiment (no need to go into the details) involving 56 students. The results were, in the authors words, tangentially supportive with p.06 I think the study was well conducted so I don't mean to slight the researchers. My guess is that if they used more subjects they probably would have reached p.05 - but would that have been an example of selective stopping? I assume it would be. So how exactly does a researcher determine beforehand - as we are suggesting they do - the number of subjects they ought to try to get for the study? I'm just not familiar with the process. Does one look at the effect sizes of previous related studies to determine if the effect is large or small and then make a decision? But let's say the effect is assumed to be small, so do you use 100 subjects? 500? How is this number determined? Appreciate the insight in this. Michael Michael A. Britt, Ph.D. mich...@thepsychfiles.commailto:mich...@thepsychfiles.com http://www.ThePsychFiles.com Twitter: @mbritt --- You are currently subscribed to tips as: pcbernha...@frostburg.edumailto:pcbernha...@frostburg.edu. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13441.4e79e96ebb5671bdb50111f18f263003n=Tl=tipso=27372 (It may be necessary to cut and paste the above URL if the line is broken) or send a blank email to leave-27372-13441.4e79e96ebb5671bdb50111f18f263...@fsulist.frostburg.edumailto:leave-27372-13441.4e79e96ebb5671bdb50111f18f263...@fsulist.frostburg.edu --- You are currently subscribed to tips as: arch...@jab.org. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5n=Tl=tipso=27373 or send a blank email to leave-27373-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
RE: [tips] Sample Size: How to Determine it?
Dear Tipsters, There are various ways to plan sample size. When teaching this in research methods, I divide the issues into two parts: 1. Estimation of population values. Here, more is better but there are diminishing returns. Think of the fact that we rarely see more than 1500 people in national polls and surveys. The formula is based on minimizing standard error. Of course, sampling is critical. 2. Conducting studies with variables: experimental, subject or correlational. There are four interconnected concepts: effect size, alpha, power and sample size. When any three are known, the fourth is determined. You can decide where to set alpha and power. For effect size (d), you can be guided by Cohen's guidelines for small, medium and large (.3, .5, .8) and choose the value you are looking for. This may come from past research or, in its absence, what you think is interesting theoretically or practically. Cohen's book on power analysis gives tables where you can look up the sample size needed after specifying the values you choose. There is also this webiste: http://homepage.stat.uiowa.edu/~rlenth/Power/ Sincerely, Stuart _ Sent via Web Access Floreat Labore Recti cultus pectora roborant Stuart J. McKelvie, Ph.D., Phone: 819 822 9600 x 2402 Department of Psychology, Fax: 819 822 9661 Bishop's University, 2600 rue College, Sherbrooke, Québec J1M 1Z7, Canada. E-mail: stuart.mckel...@ubishops.ca (or smcke...@ubishops.ca) Bishop's University Psychology Department Web Page: http://www.ubishops.ca/ccc/div/soc/psy Floreat Labore ___ From: Paul C Bernhardt [pcbernha...@frostburg.edu] Sent: 27 August 2013 08:41 To: Teaching in the Psychological Sciences (TIPS) Subject: Re: [tips] Sample Size: How to Determine it? There is software to determine this. One excellent and free app is G*Power. http://www.psycho.uni-duesseldorf.de/abteilungen/aap/gpower3/ I would use the correlational study to give me an estimate of effect size. As you describe, I would use that in the software to estimate my number of participants to attain the desired power. Practicality constraints on number of available participants usually limits things. I did such an estimate using G*Power a few weeks ago for a study we are planning. We will need to collect data over two semesters because the anticipated number of participants available from one semester's worth of students would only give us power of about .66, whereas two semester's worth would bump us up over .90. Paul On Aug 27, 2013, at 8:18 AM, Michael Britt wrote: I'm reading an interesting piece of research on anthropomorphism which essentially states after a natural disaster if we use the term mother nature when describing it, people will be less willing to contribute to relief efforts (Humanizing nature could help the perceiver to conceive natural events as imbued with intentionality and significance rather than considering them merely random and meaningless phenomena). They did two studies. Here's the issue/question: * Study 1 was correlational and involved 96 students. The results were supportive at .001 * Study 2 was an experiment (no need to go into the details) involving 56 students. The results were, in the authors words, tangentially supportive with p.06 I think the study was well conducted so I don't mean to slight the researchers. My guess is that if they used more subjects they probably would have reached p.05 - but would that have been an example of selective stopping? I assume it would be. So how exactly does a researcher determine beforehand - as we are suggesting they do - the number of subjects they ought to try to get for the study? I'm just not familiar with the process. Does one look at the effect sizes of previous related studies to determine if the effect is large or small and then make a decision? But let's say the effect is assumed to be small, so do you use 100 subjects? 500? How is this number determined? Appreciate the insight in this. Michael Michael A. Britt, Ph.D. mich...@thepsychfiles.commailto:mich...@thepsychfiles.com http://www.ThePsychFiles.com Twitter: @mbritt --- You are currently subscribed to tips as: pcbernha...@frostburg.edumailto:pcbernha...@frostburg.edu. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13441.4e79e96ebb5671bdb50111f18f263003n=Tl=tipso=27372 (It may be necessary to cut and paste the above URL if the line is broken) or send a blank email to leave-27372-13441.4e79e96ebb5671bdb50111f18f263...@fsulist.frostburg.edumailto:leave-27372-13441.4e79e96ebb5671bdb50111f18f263...@fsulist.frostburg.edu --- You are currently subscribed to tips as:
Re: [tips] Sample Size: How to Determine it?
Thanks Paul. I've downloaded G*Power. Question: the correlational component of the study revealed r = -.21, p04 (higher tendency to humanize nature were associated with a lower tendency to help victims of a natural disaster). The next test will be an independent samples t-test. How does this info help me enter the values needed by G*Power: Effect Size d and Allocation ratio N2/N1? Michael Michael A. Britt, Ph.D. mich...@thepsychfiles.com http://www.ThePsychFiles.com Twitter: @mbritt On Aug 27, 2013, at 8:41 AM, Paul C Bernhardt pcbernha...@frostburg.edu wrote: There is software to determine this. One excellent and free app is G*Power. http://www.psycho.uni-duesseldorf.de/abteilungen/aap/gpower3/ I would use the correlational study to give me an estimate of effect size. As you describe, I would use that in the software to estimate my number of participants to attain the desired power. Practicality constraints on number of available participants usually limits things. I did such an estimate using G*Power a few weeks ago for a study we are planning. We will need to collect data over two semesters because the anticipated number of participants available from one semester's worth of students would only give us power of about .66, whereas two semester's worth would bump us up over .90. Paul On Aug 27, 2013, at 8:18 AM, Michael Britt wrote: I'm reading an interesting piece of research on anthropomorphism which essentially states after a natural disaster if we use the term mother nature when describing it, people will be less willing to contribute to relief efforts (Humanizing nature could help the perceiver to conceive natural events as imbued with intentionality and significance rather than considering them merely random and meaningless phenomena). They did two studies. Here's the issue/question: Study 1 was correlational and involved 96 students. The results were supportive at .001 Study 2 was an experiment (no need to go into the details) involving 56 students. The results were, in the authors words, tangentially supportive with p.06 I think the study was well conducted so I don't mean to slight the researchers. My guess is that if they used more subjects they probably would have reached p.05 - but would that have been an example of selective stopping? I assume it would be. So how exactly does a researcher determine beforehand - as we are suggesting they do - the number of subjects they ought to try to get for the study? I'm just not familiar with the process. Does one look at the effect sizes of previous related studies to determine if the effect is large or small and then make a decision? But let's say the effect is assumed to be small, so do you use 100 subjects? 500? How is this number determined? Appreciate the insight in this. Michael Michael A. Britt, Ph.D. mich...@thepsychfiles.com http://www.ThePsychFiles.com Twitter: @mbritt --- You are currently subscribed to tips as: pcbernha...@frostburg.edu. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13441.4e79e96ebb5671bdb50111f18f263003n=Tl=tipso=27372 (It may be necessary to cut and paste the above URL if the line is broken) or send a blank email to leave-27372-13441.4e79e96ebb5671bdb50111f18f263...@fsulist.frostburg.edu --- You are currently subscribed to tips as: michael.br...@thepsychfiles.com. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13405.0125141592fa9ededc665c55d9958f69n=Tl=tipso=27373 (It may be necessary to cut and paste the above URL if the line is broken) or send a blank email to leave-27373-13405.0125141592fa9ededc665c55d9958...@fsulist.frostburg.edu --- You are currently subscribed to tips as: arch...@jab.org. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5n=Tl=tipso=27377 or send a blank email to leave-27377-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
RE: [tips] Sample Size: How to Determine it?
Hi A couple of observations to add to what others have said. First, note that the reported p (.06 or .0619, precisely) is for a non-directional test. If the authors predicted the difference, directional p is half of this (.031) and significant. This is consistent with the moderate or large value of d, depending on your preferences there. Second, I think it could be risky to use a correlational effect size to estimate an experimental effect size. In principle it could go in either direction. Experimental effect size could be larger or smaller depending on whether correlated factors in correlational study contributed to or masked the observed effect. Also would have to infer whether the experimental manipulation was more or less powerful than naturally occurring variation on the predictor/independent variable. Third, even deciding in principle what effect size was required to be important is a challenging question. The labels of small, medium, and large are pretty meaningless without knowing what generalization is being made. In this case, for example, are we trying to generalize to the helping behavior of many millions of Americans, in which case a tiny effect size could be important (as in the classic aspirin study). In reality (versus theory), research design and statistics are messy and seemingly precise tools need to be used thoughtfully. Take care Jim Jim Clark Professor Chair of Psychology 204-786-9757 4L41A From: Michael Britt [mailto:mich...@thepsychfiles.com] Also helpful. So, to answer my own previous question, based on what they found in the correlational study and what one might guess from previous research, I'm going to assume that the effect size here, if it exists, is probably small. So I used .3 in G*Power. The result? G*Power suggests that I get 242 subjects per group. These researchers had 26 subjects in each group. So: if you were the reviewer what would you conclude? The researchers found: ...the results revealed that participants in the anthropomorphism condition were tendentially less willing to help the victims of the natural disaster (M = 4.39, SD = 1.02) than participants in the control condition (M = 4.89, SD = 0.87), t(50) = -1.91, p = .06, d = 0.53. Would you recommend that they get more subjects? Michael Michael A. Britt, Ph.D. mich...@thepsychfiles.commailto:mich...@thepsychfiles.com http://www.ThePsychFiles.com Twitter: @mbritt On Aug 27, 2013, at 8:59 AM, Stuart McKelvie smcke...@ubishops.camailto:smcke...@ubishops.ca wrote: Dear Tipsters, There are various ways to plan sample size. When teaching this in research methods, I divide the issues into two parts: 1. Estimation of population values. Here, more is better but there are diminishing returns. Think of the fact that we rarely see more than 1500 people in national polls and surveys. The formula is based on minimizing standard error. Of course, sampling is critical. 2. Conducting studies with variables: experimental, subject or correlational. There are four interconnected concepts: effect size, alpha, power and sample size. When any three are known, the fourth is determined. You can decide where to set alpha and power. For effect size (d), you can be guided by Cohen's guidelines for small, medium and large (.3, .5, .8) and choose the value you are looking for. This may come from past research or, in its absence, what you think is interesting theoretically or practically. Cohen's book on power analysis gives tables where you can look up the sample size needed after specifying the values you choose. There is also this webiste: http://homepage.stat.uiowa.edu/~rlenth/Power/ Sincerely, Stuart _ Sent via Web Access Floreat Labore Recti cultus pectora roborant Stuart J. McKelvie, Ph.D., Phone: 819 822 9600 x 2402 Department of Psychology, Fax: 819 822 9661 Bishop's University, 2600 rue College, Sherbrooke, Québec J1M 1Z7, Canada. E-mail: stuart.mckel...@ubishops.camailto:stuart.mckel...@ubishops.ca (or smcke...@ubishops.camailto:smcke...@ubishops.ca) Bishop's University Psychology Department Web Page: http://www.ubishops.ca/ccc/div/soc/psy Floreat Labore ___ From: Paul C Bernhardt [pcbernha...@frostburg.edumailto:pcbernha...@frostburg.edu] Sent: 27 August 2013 08:41 To: Teaching in the Psychological Sciences (TIPS) Subject: Re: [tips] Sample Size: How to Determine it? There is software to determine this. One excellent and free app is G*Power. http://www.psycho.uni-duesseldorf.de/abteilungen/aap/gpower3/ I would use the correlational study to give me an estimate of effect size. As you describe, I would use that in the software to estimate my number of participants to attain the desired
Re: [tips] Sample Size: How to Determine it?
Hi Michael: Be careful with the effect size statistic that G*Power uses, sometimes it is using rho. rho = .3 would be a medium effect size. Ken PS - It is surprising how underpowered are many of the experiments reported in the journals. Kenneth M. Steele, Ph. D.steel...@appstate.edu Professor Department of Psychology http://www.psych.appstate.edu Appalachian State University Boone, NC 28608 USA On 8/27/2013 9:59 AM, Michael Britt wrote: Also helpful. So, to answer my own previous question, based on what they found in the correlational study and what one might guess from previous research, I'm going to assume that the effect size here, if it exists, is probably small. So I used .3 in G*Power. The result? G*Power suggests that I get 242 subjects per group. These researchers had 26 subjects in each group. So: if you were the reviewer what would you conclude? The researchers found: ...the results revealed that participants in the anthropomorphism condition were tendentially less willing to help the victims of the natural disaster (M = 4.39, SD = 1.02) than participants in the control condition (M = 4.89, SD = 0.87), t(50) = –1.91, p = .06, d = 0.53. Would you recommend that they get more subjects? Michael Michael A. Britt, Ph.D. mich...@thepsychfiles.com mailto:mich...@thepsychfiles.com http://www.ThePsychFiles.com Twitter: @mbritt On Aug 27, 2013, at 8:59 AM, Stuart McKelvie smcke...@ubishops.ca mailto:smcke...@ubishops.ca wrote: Dear Tipsters, There are various ways to plan sample size. When teaching this in research methods, I divide the issues into two parts: 1. Estimation of population values. Here, more is better but there are diminishing returns. Think of the fact that we rarely see more than 1500 people in national polls and surveys. The formula is based on minimizing standard error. Of course, sampling is critical. 2. Conducting studies with variables: experimental, subject or correlational. There are four interconnected concepts: effect size, alpha, power and sample size. When any three are known, the fourth is determined. You can decide where to set alpha and power. For effect size (d), you can be guided by Cohen's guidelines for small, medium and large (.3, .5, .8) and choose the value you are looking for. This may come from past research or, in its absence, what you think is interesting theoretically or practically. Cohen's book on power analysis gives tables where you can look up the sample size needed after specifying the values you choose. There is also this webiste: http://homepage.stat.uiowa.edu/~rlenth/Power/ Sincerely, Stuart _ Sent via Web Access Floreat Labore Recti cultus pectora roborant Stuart J. McKelvie, Ph.D., Phone: 819 822 9600 x 2402 Department of Psychology, Fax: 819 822 9661 Bishop's University, 2600 rue College, Sherbrooke, Québec J1M 1Z7, Canada. E-mail: stuart.mckel...@ubishops.ca mailto:stuart.mckel...@ubishops.ca (or smcke...@ubishops.ca mailto:smcke...@ubishops.ca) Bishop's University Psychology Department Web Page: http://www.ubishops.ca/ccc/div/soc/psy Floreat Labore ___ *From:*Paul C Bernhardt [pcbernha...@frostburg.edu mailto:pcbernha...@frostburg.edu] *Sent:*27 August 2013 08:41 *To:*Teaching in the Psychological Sciences (TIPS) *Subject:*Re: [tips] Sample Size: How to Determine it? There is software to determine this. One excellent and free app is G*Power. http://www.psycho.uni-duesseldorf.de/abteilungen/aap/gpower3/ I would use the correlational study to give me an estimate of effect size. As you describe, I would use that in the software to estimate my number of participants to attain the desired power. Practicality constraints on number of available participants usually limits things. I did such an estimate using G*Power a few weeks ago for a study we are planning. We will need to collect data over two semesters because the anticipated number of participants available from one semester's worth of students would only give us power of about .66, whereas two semester's worth would bump us up over .90. Paul On Aug 27, 2013, at 8:18 AM, Michael Britt wrote: I'm reading an interesting piece of research on anthropomorphism which essentially states after a natural disaster if we use the term mother nature when describing it, people will be less willing to contribute to relief efforts (Humanizing nature could help the perceiver to conceive natural events as imbued with intentionality and significance rather than considering them merely random and meaningless phenomena). They did two studies. Here's the issue/question: * Study 1 was correlational and
RE: [tips] Sample Size: How to Determine it?
I am assuming this was an independent samples t test where some participants heard the mother nature language and others didn't. Using the d of .53 they obtained as my estimate of what effect size they would be interested in obtaining (or that they think would be worthwhile to note), it appears that, with a df of 50, they had less than a 50/50 chance of finding a significant result of that size if one existed in the population. As others have pointed out, you need to determine before the study begins, what effect size you are interested in obtaining. For example, you may believe that even a .05 effect size (1/20th of a standard deviation difference between the two means) could be meaningful given the question. If so, you are going to need a very large sample size to have a high probability of finding a significant result if such a small difference exists in the population. By my calculations*, if you wanted to have at least an 80 percent chance of detecting an effect size of at least .50 (half a standard deviation difference between the means) with an independent sample t test, you would need to have 128 participants in the study (64 in each group). If you wanted to have an 80% chance of detecting a .05 (5 percent) effect size in such a case, you would need 12560 participants (6280 in each group). *My power calculations came from http://homepage.stat.uiowa.edu/~rlenth/Power/. The author has a nice discussion of power and why retrospective power analysis is worthless under the Advice section on that page. Rick Dr. Rick Froman, Chair Division of Humanities and Social Sciences Box 3519 x7295 rfro...@jbu.edumailto:rfro...@jbu.edu http://bit.ly/DrFroman Proverbs 14:15 A simple man believes anything, but a prudent man gives thought to his steps. From: Michael Britt [mailto:mich...@thepsychfiles.com] Sent: Tuesday, August 27, 2013 9:00 AM To: Teaching in the Psychological Sciences (TIPS) Subject: Re: [tips] Sample Size: How to Determine it? Also helpful. So, to answer my own previous question, based on what they found in the correlational study and what one might guess from previous research, I'm going to assume that the effect size here, if it exists, is probably small. So I used .3 in G*Power. The result? G*Power suggests that I get 242 subjects per group. These researchers had 26 subjects in each group. So: if you were the reviewer what would you conclude? The researchers found: ...the results revealed that participants in the anthropomorphism condition were tendentially less willing to help the victims of the natural disaster (M = 4.39, SD = 1.02) than participants in the control condition (M = 4.89, SD = 0.87), t(50) = -1.91, p = .06, d = 0.53. Would you recommend that they get more subjects? Michael Michael A. Britt, Ph.D. mich...@thepsychfiles.commailto:mich...@thepsychfiles.com http://www.ThePsychFiles.com Twitter: @mbritt On Aug 27, 2013, at 8:59 AM, Stuart McKelvie smcke...@ubishops.camailto:smcke...@ubishops.ca wrote: Dear Tipsters, There are various ways to plan sample size. When teaching this in research methods, I divide the issues into two parts: 1. Estimation of population values. Here, more is better but there are diminishing returns. Think of the fact that we rarely see more than 1500 people in national polls and surveys. The formula is based on minimizing standard error. Of course, sampling is critical. 2. Conducting studies with variables: experimental, subject or correlational. There are four interconnected concepts: effect size, alpha, power and sample size. When any three are known, the fourth is determined. You can decide where to set alpha and power. For effect size (d), you can be guided by Cohen's guidelines for small, medium and large (.3, .5, .8) and choose the value you are looking for. This may come from past research or, in its absence, what you think is interesting theoretically or practically. Cohen's book on power analysis gives tables where you can look up the sample size needed after specifying the values you choose. There is also this webiste: http://homepage.stat.uiowa.edu/~rlenth/Power/ Sincerely, Stuart _ Sent via Web Access Floreat Labore Recti cultus pectora roborant Stuart J. McKelvie, Ph.D., Phone: 819 822 9600 x 2402 Department of Psychology, Fax: 819 822 9661 Bishop's University, 2600 rue College, Sherbrooke, Québec J1M 1Z7, Canada. E-mail: stuart.mckel...@ubishops.camailto:stuart.mckel...@ubishops.ca (or smcke...@ubishops.camailto:smcke...@ubishops.ca) Bishop's University Psychology Department Web Page: http://www.ubishops.ca/ccc/div/soc/psy Floreat Labore ___ From: Paul C Bernhardt
RE: [tips] Sample Size: How to Determine it?
I was going to stay out of this discussion but I have to address a couple of points, one of which is made by Rick at the end of his post: (1) The major problem with power analysis is that it requires one to have knowledge of POPULATION PARAMETERS, that is, the means, the standard deviation, the correlations, and so on. NOTE: a researcher has sample data from which descriptive statistics and inferential statistics are calculated which will have sampling error and possible other types of error that make the sample estimates of the mean, standard deviation, correlation, etc., misleading. The proper thing to do before collecting the data is to conduct an A Priori power analysis. But An A Priori power analysis assumes that one knows the relevant population means, standard deviations, correlations, effect size, and so on that are involved. This is a problem because far too many researchers don't have a clue what these values are or should be. If you don't know what the population parameters are, step away from the data and let a professional try to do something with it. (2) Rick Froman below refers to Russ Lenth's website where one can use his software for some calculations -- I suggest one use G*Power instead -- as well as his position that retrospective or observed power analysis is bad, m'kay? I suggest that one instead read Geoff Cumming's Understanding the New Statistics which goes into much more detail about effect sizes, confidence intervals, and meta-analysis -- all of which are inter-related; see: http://www.amazon.com/Understanding-The-New-Statistics-Meta-Analysis/dp/041587968X/ref=sr_1_1?ie=UTF8qid=1377628036sr=8-1keywords=cummings+meta-analysis Cummings makes a stronger argument than Lenth. However, I would also suggest that one read my review of Cummings' book in PsycCritiques which takes issue with the anti-retrospective or anti-observed power analysis situation; see: Palij, M. (2012). New statistical rituals for old. PsycCRITIQUES 57 (24). (3) Pragmatically, most psychologists who do statistical analysis rely almost solely on the sample information to reach conclusions about the population parameters. This is where concerns about whether the probability of one's obtained statistic like a t-test is statistically significant or what to do if one has a p(obt t)= .06. The p-value doesn't really matter if you know that that two sample means you have come from different populations, right? Which is why one is urged to use confidence intervals instead. But psychologists will look at the observed power level provided by SPSS' MANOVA or GLM procedures if they have done an ANOVA because they did not select the power level before they collected their data. And it is only then that they might realize, Ooops!, I don't really have enough statistical power to reject a false null hypothesis. But this is an old tale that all Tipsters should be familiar with given our current statistical practices -- see Cummings' book if one needs a refresher on what some consider proper statistical analysis in contemporary psychological research. Then, again, really knowing the phenomenon you're studying and having strong theory, such as signal detection theory in psychophysics or recognition memory research, may go a much longer way than wondering whether one has a statistically significant result. -Mike Palij New York University m...@nyu.edu Original Message On Tue, 27 Aug 2013 10:53:07 -0700, Rick Froman wrote: I am assuming this was an independent samples t test where some participants heard the mother nature language and others didn't. Using the d of .53 they obtained as my estimate of what effect size they would be interested in obtaining (or that they think would be worthwhile to note), it appears that, with a df of 50, they had less than a 50/50 chance of finding a significant result of that size if one existed in the population. As others have pointed out, you need to determine before the study begins, what effect size you are interested in obtaining. For example, you may believe that even a .05 effect size (1/20th of a standard deviation difference between the two means) could be meaningful given the question. If so, you are going to need a very large sample size to have a high probability of finding a significant result if such a small difference exists in the population. By my calculations*, if you wanted to have at least an 80 percent chance of detecting an effect size of at least .50 (half a standard deviation difference between the means) with an independent sample t test, you would need to have 128 participants in the study (64 in each group). If you wanted to have an 80% chance of detecting a .05 (5 percent) effect size in such a case, you would need 12560 participants (6280 in each group). *My power calculations came from http://homepage.stat.uiowa.edu/~rlenth/Power/ . The author has a nice discussion of power and why retrospective power
[tips] Tenure Track Opening: Adult Clinical
West Chester University invites applications for a tenure-track faculty position in Clinical Psychology at the assistant professor level. Applicants must have a Ph.D. in psychology, an active program of research with an adult population, and the ability to mentor undergraduate and graduate students in research. We are especially interested in recruiting faculty from under-represented groups. Preference will be given to candidates who can begin their appointment in January, 2014. Applicants must express a commitment to teaching courses in clinical psychology (e.g., personality, abnormal/psychopathology, counseling/psychotherapy, testing/assessment) at the undergraduate and graduate levels. Applicants selected for on-campus interviews will present a colloquium to demonstrate teaching and research excellence. The Department of Psychology is composed of 20 full-time faculty members and serves approximately 800 undergraduate majors and 100 master’s degree students. Additional information is available at http://www.wcupa.edu. Completion of the Ph.D. from an APA-approved program is required before the start of the appointment, as is licensure-eligibility in the state of PA. Applicants should apply online at https://wcupa.peopleadmin.com. Applications must include a letter identifying the courses the applicant is prepared to teach, curriculum vitae, 3 letters of reference, no more than 3 reprints or preprints of published articles, a statement of teaching philosophy, and a statement of research interests. Review of candidates will begin on August 15, 2013 and continue until the position is filled. Applicants must successfully complete the interview process and a colloquium to be considered finalists. The filling of this position is contingent upon available funding. All offers of employment are subject to and contingent upon satisfactory completion of all pre-employment criminal background and consumer reporting checks. West Chester University is an Affirmative Action-Equal Opportunity Employer. --- You are currently subscribed to tips as: arch...@jab.org. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5n=Tl=tipso=27389 or send a blank email to leave-27389-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
RE: [tips] Sample Size: How to Determine it?
My two cents: Decide on what is the smallest effect that you would consider to be of importance. If you think Type I and Type II errors are equally serious, then set both alpha and beta to .05, that is, find N for 95% power. G*Power does this with ease, but you are unlikely to like the answer. If precision of estimation of the effect size is of importance, even bigger is better (less wide confidence intervals for effect size). With respect to independent samples t test, you can use the procedure that G*Power specifies for that design, with d as the effect size, or you can use the point biserial regression procedure, with r as the effect size. Do note that the size of the point biserial is greatly affected by the ratio of the sample sizes, which is not true with d. See http://core.ecu.edu/psyc/wuenschk/StatHelp/d-r.htm When discussing issues of power and effect size, always pay attention to speakers from NYU. :-) Cheers, Karl L. Wuensch -Original Message- From: Mike Palij [mailto:m...@nyu.edu] Sent: Tuesday, August 27, 2013 3:06 PM To: Teaching in the Psychological Sciences (TIPS) Cc: Michael Palij Subject: RE: [tips] Sample Size: How to Determine it? I was going to stay out of this discussion but I have to address a couple of points, one of which is made by Rick at the end of his post: (1) The major problem with power analysis is that it requires one to have knowledge of POPULATION PARAMETERS, that is, the means, the standard deviation, the correlations, and so on. NOTE: a researcher has sample data from which descriptive statistics and inferential statistics are calculated which will have sampling error and possible other types of error that make the sample estimates of the mean, standard deviation, correlation, etc., misleading. The proper thing to do before collecting the data is to conduct an A Priori power analysis. But An A Priori power analysis assumes that one knows the relevant population means, standard deviations, correlations, effect size, and so on that are involved. This is a problem because far too many researchers don't have a clue what these values are or should be. If you don't know what the population parameters are, step away from the data and let a professional try to do something with it. (2) Rick Froman below refers to Russ Lenth's website where one can use his software for some calculations -- I suggest one use G*Power instead -- as well as his position that retrospective or observed power analysis is bad, m'kay? I suggest that one instead read Geoff Cumming's Understanding the New Statistics which goes into much more detail about effect sizes, confidence intervals, and meta-analysis -- all of which are inter-related; see: http://www.amazon.com/Understanding-The-New-Statistics-Meta-Analysis/dp/041587968X/ref=sr_1_1?ie=UTF8qid=1377628036sr=8-1keywords=cummings+meta-analysis Cummings makes a stronger argument than Lenth. However, I would also suggest that one read my review of Cummings' book in PsycCritiques which takes issue with the anti-retrospective or anti-observed power analysis situation; see: Palij, M. (2012). New statistical rituals for old. PsycCRITIQUES 57 (24). (3) Pragmatically, most psychologists who do statistical analysis rely almost solely on the sample information to reach conclusions about the population parameters. This is where concerns about whether the probability of one's obtained statistic like a t-test is statistically significant or what to do if one has a p(obt t)= .06. The p-value doesn't really matter if you know that that two sample means you have come from different populations, right? Which is why one is urged to use confidence intervals instead. But psychologists will look at the observed power level provided by SPSS' MANOVA or GLM procedures if they have done an ANOVA because they did not select the power level before they collected their data. And it is only then that they might realize, Ooops!, I don't really have enough statistical power to reject a false null hypothesis. But this is an old tale that all Tipsters should be familiar with given our current statistical practices -- see Cummings' book if one needs a refresher on what some consider proper statistical analysis in contemporary psychological research. Then, again, really knowing the phenomenon you're studying and having strong theory, such as signal detection theory in psychophysics or recognition memory research, may go a much longer way than wondering whether one has a statistically significant result. -Mike Palij New York University m...@nyu.edu Original Message On Tue, 27 Aug 2013 10:53:07 -0700, Rick Froman wrote: I am assuming this was an independent samples t test where some participants heard the mother nature language and others didn't. Using the d of .53 they obtained as my estimate of what effect size they would be
[tips] noninvasive brain-to-brain interface in human beings
Another interesting tidbit from Science Daily: http://www.sciencedaily.com/releases/2013/08/130827122713.htm -- Carol DeVolder, Ph.D. Professor of Psychology St. Ambrose University 518 West Locust Street Davenport, Iowa 52803 563-333-6482 --- You are currently subscribed to tips as: arch...@jab.org. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5n=Tl=tipso=27395 or send a blank email to leave-27395-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
Re: [tips] Sample Size: How to Determine it?
It's hard to know what I would conclude. Are there other significant effects found in the study with the small sample size? Power is not an issue if you have statistical significance. I have a published paper in which one study has an N of 8, four in each of the two groups. It was significant with a huge effect size (naturally, with such a small sample). People may not believe it (is it robust?) but power is not the reason to doubt it. With a marginally significant effect (I've never heard the term tendentially used in this context) it is essentially the same problem. It is what it is. They don't have statistical significance. They have a small, but not absurdly small, sample size (how many studies with df = 50 are published out there?). I would be less concerned with the stats on this and more concerned with the claims they make in their discussion about the finding. Do they write their discussion with scant or no real accounting for the fact that it is marginally significant? They need to describe their finding as tentative and suggestive that a future study needs greater control over the IV and possibly increased sample size. If they try to act like they have a big discovery, I'd be requesting a rewrite. In fact, an important part of the interpretation is the degree of surprise in the finding. Is it consistent with other findings in the domain? If so, then they can speak more strongly (not a lot, just a little) about the meaning of their findings. If it is surprising and contrary to other findings in the literature, then I'd be prone to rejection of the article due to lack of a sufficient finding to change my prior view. Paul On Aug 27, 2013, at 9:59 AM, Michael Britt wrote: Also helpful. So, to answer my own previous question, based on what they found in the correlational study and what one might guess from previous research, I'm going to assume that the effect size here, if it exists, is probably small. So I used .3 in G*Power. The result? G*Power suggests that I get 242 subjects per group. These researchers had 26 subjects in each group. So: if you were the reviewer what would you conclude? The researchers found: ...the results revealed that participants in the anthropomorphism condition were tendentially less willing to help the victims of the natural disaster (M = 4.39, SD = 1.02) than participants in the control condition (M = 4.89, SD = 0.87), t(50) = –1.91, p = .06, d = 0.53. Would you recommend that they get more subjects? Michael Michael A. Britt, Ph.D. mich...@thepsychfiles.commailto:mich...@thepsychfiles.com http://www.ThePsychFiles.com Twitter: @mbritt On Aug 27, 2013, at 8:59 AM, Stuart McKelvie smcke...@ubishops.camailto:smcke...@ubishops.ca wrote: Dear Tipsters, There are various ways to plan sample size. When teaching this in research methods, I divide the issues into two parts: 1. Estimation of population values. Here, more is better but there are diminishing returns. Think of the fact that we rarely see more than 1500 people in national polls and surveys. The formula is based on minimizing standard error. Of course, sampling is critical. 2. Conducting studies with variables: experimental, subject or correlational. There are four interconnected concepts: effect size, alpha, power and sample size. When any three are known, the fourth is determined. You can decide where to set alpha and power. For effect size (d), you can be guided by Cohen's guidelines for small, medium and large (.3, .5, .8) and choose the value you are looking for. This may come from past research or, in its absence, what you think is interesting theoretically or practically. Cohen's book on power analysis gives tables where you can look up the sample size needed after specifying the values you choose. There is also this webiste: http://homepage.stat.uiowa.edu/~rlenth/Power/ Sincerely, Stuart _ Sent via Web Access Floreat Labore Recti cultus pectora roborant Stuart J. McKelvie, Ph.D., Phone: 819 822 9600 x 2402 Department of Psychology, Fax: 819 822 9661 Bishop's University, 2600 rue College, Sherbrooke, Québec J1M 1Z7, Canada. E-mail: stuart.mckel...@ubishops.camailto:stuart.mckel...@ubishops.ca (or smcke...@ubishops.camailto:smcke...@ubishops.ca) Bishop's University Psychology Department Web Page: http://www.ubishops.ca/ccc/div/soc/psy Floreat Labore ___ From: Paul C Bernhardt [pcbernha...@frostburg.edumailto:pcbernha...@frostburg.edu] Sent: 27 August 2013 08:41 To: Teaching in the Psychological Sciences (TIPS) Subject: Re: [tips] Sample Size: How to Determine it? There is software to determine this. One excellent and free app is G*Power.
[tips] Belfast not Dublin,you dummy.
On tonight's Jeopardy game the final question dealt with what area of Ireland are there distinctive markings indicating Protestant and Catholic neighborhoods.All three of the contestants wrote down Dublin.I wonder if they have heard of William of Orange. Btw,I hosted a college student from Northern Ireland a fortnight ago.She told me that when they play the Monopoly board game,players avoid landing on Dublin. Apparently Cork is esteemed as the real capiyal of the Irish Republic. michael --- You are currently subscribed to tips as: arch...@jab.org. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5n=Tl=tipso=27398 or send a blank email to leave-27398-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu
Fw: [tips] Tenure Track Opening:
Donald Trump University Please submit $4000 with youur application michael --- You are currently subscribed to tips as: arch...@jab.org. To unsubscribe click here: http://fsulist.frostburg.edu/u?id=13090.68da6e6e5325aa33287ff385b70df5d5n=Tl=tipso=27399 or send a blank email to leave-27399-13090.68da6e6e5325aa33287ff385b70df...@fsulist.frostburg.edu