yes i found how to do the pvalue and fold change. I am using BRB array tools for my analysis and since it is a freeware, for more than 2 time points, i need to manually get the fold change based on the intensity values. But the pvalues are listed. I wanted to know if the pvalue is associated to the gene alone or is it associated to the fold change also.
One needs to be very cautious about the meaning of p-values in microarray experiments. The p-value for a fold change measures the probability that that change is unlikely to be due to chance (random error). However, if an array has 32,000 genes in it, all of which can be compared for change, then for a p
Mr CHRISTOPHER has given the correct explanation of p-value. For any observation made, p-value of 0.05 is taken as the minimum level of statistical significance,meaning thereby that the chance of the observation made being false positive is 1:20 [null hypothesis].
Thanks a lot for helping me. This information is really useful.
@Christopher- I would go through Stephen Chu group's publications and get some information from it too.
is it that the pvalue which we compute changes according to the time point. Because from what I observe from the results I got from BRB array tool, the p value is computed for each gene in the statistic over certain number of permutations and it does not change according to the time point. But fold change on the other hand changes according to time point. Like if we compare same set of n number of genes for 6hr, drug vs no drug treatment, then the fold change is different than when we compare 12hr drug vs no drug.
@Aparna - For each comaprison that you make you can estimate a p-value. Thus the difference between 12 hr drug vs no drug is a different question from asking if there is a difference at 6 hours or if 6 genes are all different (drug vs no drug) at either time. It is a question of what are the individual mean values and their precision estimates (standard deviation or standard error, depending on whether the desired prediction is for the next data point or the next repetition of the experiment) and how far apart they are and thus how unlikely is it that this result could be due to random chance. Of course, if you take p
I am not familair with the specific BRB Array tools, but the statistical question is quite simple. Unless you a priori limit yourself to only a few comparisons, then the ability to achieve 95% confidence of a difference (p
Yeah maybe I would look through the chapter and see how best I can deal with my data. I was also thinking of applying teh Fischer test or student t test and calculate the p value. Since the tool I am working with is already stringent and gives me very few genes as significant, using bonferroni correction would further decrease the number of genes. I shall work on it and let you know which suited the best.
Both the Fisher (F) test and the Student's t test need to be handled in a way that a large number of possible comparisons does not guarantee that there will be many false positives. The Student-Neuman-Keuls test is really a Student's t test taking into account the number of tests that could be made and therefore looking at the correct level in the t-table for p-values. Any one test will give you a result, but it is the combination of individual tests that causes the logical error of thinking that what is significant in one test is still significant when there are many tests. Confirming your array result with an RT PCR assay will give you the Gold Standard for is the difference real. The differences should be so large that you don't need statistics to see the difference. (Statistics is the tool for determining the probability that two (or more) readings or means are not the same. When the difference is very large compared to the scatter of data or mean values (as indicated by std dev or sem) then you don't need a statistical test to tell you that the probability of their being from the same population is very small and hence unlikely.
'/I think I will need to focus a lot more in this area to understand some of the concepts and explanations you have given me since I am new to the field and have been working on micro