tag:blogger.com,1999:blog-5547907074344788039.post5633805100742730363..comments2016-09-08T05:37:41.761-07:00Comments on p-value.info: What's the significance of 0.05 significance?Carl Andersonhttp://www.blogger.com/profile/11930448254473684406noreply@blogger.comBlogger14125tag:blogger.com,1999:blog-5547907074344788039.post-7556207518069059612016-07-30T09:57:51.149-07:002016-07-30T09:57:51.149-07:00How can I select P>0.05, p>0.01 and p>0.0...How can I select P>0.05, p>0.01 and p>0.001<br />fouzia afzalhttps://www.blogger.com/profile/07118686963264350208noreply@blogger.comtag:blogger.com,1999:blog-5547907074344788039.post-28235862426574499352016-07-30T09:57:08.088-07:002016-07-30T09:57:08.088-07:00How can I select P>0.05, p>0.01 and p>0.0...How can I select P>0.05, p>0.01 and p>0.001<br />Unknownhttps://www.blogger.com/profile/07118686963264350208noreply@blogger.comtag:blogger.com,1999:blog-5547907074344788039.post-43433782225248373962016-05-31T06:52:41.728-07:002016-05-31T06:52:41.728-07:00Carl,
Need your help looking over this data.
For ...Carl, <br />Need your help looking over this data.<br />For 1st condition<br />OR = 1.57 <br />95% CI = (0.40, 6.33)<br />For 2nd condition<br />OR = 1.35 <br />95% CI = (0.20, 8.86)<br />Both 95% confidence intervals contain 1, thus are not statistically significant at alpha = 0.05.<br />So I accept the null, right?<br />I'm a bit rusty with my stats and am having a little bit of a hard time. <br /><br />Diana Dhttps://www.blogger.com/profile/05037645221673446254noreply@blogger.comtag:blogger.com,1999:blog-5547907074344788039.post-23995691017065358912015-10-24T04:11:09.772-07:002015-10-24T04:11:09.772-07:00John, your null hypothesis simply states that ther...John, your null hypothesis simply states that there is no difference. It doesn't say anything about significance. You are trying to find evidence to reject it. You found a likelihood of 0.05 of seeing the differences you found, or more extreme, if the null hypothesis were true. Because probability is continuous, p < 0.05 is essentially the same as p <= 0.05, so you can consider it significant at 5% level and reject the null hypothesis.Carl Andersonhttps://www.blogger.com/profile/11930448254473684406noreply@blogger.comtag:blogger.com,1999:blog-5547907074344788039.post-53158433189435360442015-10-23T19:27:34.865-07:002015-10-23T19:27:34.865-07:00Hi Carl. I had a p value of exactly 0.05. how will...Hi Carl. I had a p value of exactly 0.05. how will i make my conclusion when my null hypothesis says there is no significant difference?John Maryhttps://www.blogger.com/profile/05332219450532265884noreply@blogger.comtag:blogger.com,1999:blog-5547907074344788039.post-12562861022523215622015-09-20T05:51:49.523-07:002015-09-20T05:51:49.523-07:00Thankyou very much Carl,your simple explanation he...Thankyou very much Carl,your simple explanation helped me a lot . I am new to all this and struggling to understand.<br />moona sarwathttps://www.blogger.com/profile/12849320468743159687noreply@blogger.comtag:blogger.com,1999:blog-5547907074344788039.post-17230179808653250282015-09-20T04:17:47.997-07:002015-09-20T04:17:47.997-07:00Mohd,
See my response to the comment below. Hopefu...Mohd,<br />See my response to the comment below. Hopefully that helps clarify things for you.Carl Andersonhttps://www.blogger.com/profile/11930448254473684406noreply@blogger.comtag:blogger.com,1999:blog-5547907074344788039.post-9909630808013662382015-09-20T04:16:23.621-07:002015-09-20T04:16:23.621-07:00Moona,
I don't fully understand what you are t...Moona,<br />I don't fully understand what you are trying to do but the answer is no. You can't add p-values. <br />There is some muddled thinking here I would like to clear up. P-values are completely independent of significance level. Significance is layer of interpretation after a p-value is obtained.<br />Here is the flow:<br />1) Set up null hypothesis: metric_control = metric_treatment<br />2) set up alternative: metric_control != metric_treament<br />3) compute test statistic: say t = 3.95<br />4) statistical software will work out p-values associate with that test statistic (for the level of degrees of freedom). Say, p=0.03<br />That p-value is the probability of obtaining the metric value, or more extreme, if the null hypothesis were true.<br />Notice that I haven't mentioned significance yet. We got a probability of 0.03, a 3% chance that we would get these results (or more extreme) by chance if the null hypothesis were true. <br />5) Now you have to interpret how strong a signal that is. If you chose 5% significance level, as 0.03 < 0.05, it is significant at 5%. If you choose 1% significance level, then as 0.03 > 0.01 then it is not significant. A significance level is some critical threshold for our p-value used. A p-value of 0.03 is significant at 3% level and it is significant at 4% level and at 5% level, ... and at 99% level.<br /><br />The p-value is a probability associated with a given null hypothesis. As such, you can't sum them across hypotheses like this.<br /><br />(In Bayesian statistics, however, you work with likelihood of events which can be multiplied to get joint probabilities but that is very different statistical approach.)<br />Carl Andersonhttps://www.blogger.com/profile/11930448254473684406noreply@blogger.comtag:blogger.com,1999:blog-5547907074344788039.post-13054335191544268162015-09-19T23:37:29.287-07:002015-09-19T23:37:29.287-07:00Hi there Carl,
I am trying to write up a project ...Hi there Carl,<br />I am trying to write up a project proposal for the very first time , I am given, 'In-silico studies of metasignature genes in Lung Cancer'.<br />After log transformation and student t test, p values are obtained at the significance fo 0.05.<br />what I would like to know whether we could sum the p-values obtained from using significance level of 0.01,then again using the same set of genes and setting the significance at 0.02 thus calculatiing till 0.05, and then adjusting the p-values using FDR.<br />will these produce more significant differentially expressed genes? moona sarwathttps://www.blogger.com/profile/12849320468743159687noreply@blogger.comtag:blogger.com,1999:blog-5547907074344788039.post-81157732171395714292015-09-13T20:21:18.189-07:002015-09-13T20:21:18.189-07:00Could you explain why 0.05 is used to limit the p-...Could you explain why 0.05 is used to limit the p-value ?Mohd Hafizihttps://www.blogger.com/profile/06937018402623028756noreply@blogger.comtag:blogger.com,1999:blog-5547907074344788039.post-72229101367027176362015-05-13T20:09:38.209-07:002015-05-13T20:09:38.209-07:00Lukey,
Conceptually, the F statistic is the ratio...Lukey,<br /><br />Conceptually, the F statistic is the ratio of the variance explained by a treatment or independent variable / variance explained by noise. Essentially it is a signal to noise ratio. <br />The higher F, the stronger the signal relative to the noise or error term. Strong signals will typically have much higher variance from treatment than error term and those treatment variances will be much larger than the F value. For instance: if F = signal / noise = 100 / 20 = 5 and so 100 is much greater than 5. <br /><br />You determine significance not by comparing standard deviation and F but simply by looking F up in a lookup table and it returns a p-value. Of course, all statistical software will both compute F and provide the p-value in their output.<br /><br />CarlCarl Andersonhttps://www.blogger.com/profile/11930448254473684406noreply@blogger.comtag:blogger.com,1999:blog-5547907074344788039.post-46486358495778324242015-05-07T04:19:17.585-07:002015-05-07T04:19:17.585-07:00what conclusions can be drawn about statistical si...what conclusions can be drawn about statistical significance when the standard deviation is greater than the F value?Lukey Mabunghttps://www.blogger.com/profile/16215459051867706187noreply@blogger.comtag:blogger.com,1999:blog-5547907074344788039.post-22530597989649770832013-09-27T06:43:02.733-07:002013-09-27T06:43:02.733-07:00That really depends on your null hypothesis and ex...That really depends on your null hypothesis and experimental design. Suppose I have one group of people sit in a chair for 5 mins and another run for 5 mins and then I measure their pulse. I find that the runners have a significantly higher pulse. Does that provide any meaningful insights? No, not really. If instead, I give the treatment group a new drug, well that is likely different. Sorry, but it is hard to give a good answer to this because it is very much tied up with the basis of statistical inference. I would suggest reading up on that, starting with the null hypothesis.Carl Andersonhttps://www.blogger.com/profile/11930448254473684406noreply@blogger.comtag:blogger.com,1999:blog-5547907074344788039.post-53474779117096723142013-09-23T20:54:31.539-07:002013-09-23T20:54:31.539-07:00Carl,
I am trying to understand p<0.01 to answe...Carl,<br />I am trying to understand p<0.01 to answer the following: A hypothetical experimental clinical research study found a significant difference between the results for the treatment group and results for the control group (p<.01). <br /><br />Should we, as consumers of research, have confidence that the statistically significant findings are also clinically significant? What kinds of questions might we want to consider before we can answer that question? <br /><br />Can you help me or point me to see the "light"? Thank you in advance.Guardianhttps://www.blogger.com/profile/04320859392210931607noreply@blogger.com