Tell me more ×
Cross Validated is a question and answer site for statisticians, data analysts, data miners and data visualization experts. It's 100% free, no registration required.
up vote 3 down vote favorite
1

I would like to confirm my biological hypothesis with statistical approach.

Red line on picture represents average frequency of some biological phenomenom for whole chromosome. Chromosome consists of milions of bases, which can be understood as positions. So that we have milions of positions total. Then I, using certain rule, pick up 100 positions. For these 100 positions I determine frequency of the same biological phenomenon (so that I have 100 isolated points, for every x there is y and they are connected by lines, as seen on picture). Because picture shows just these positions, red line does not represent average of picture (it represents average of whole chromosome).

I want to show that freguency at my certain positions is significantly higher than average (so that the biological process I observe is really there).

My data are from database,so my output is what I calculated using this database-there are no more measurements or data of this kind.

y-axis represent counts, not ratios. Adjacent points on graph are neighbours on the chromosome, so one could expect that two neighbours on the chromosomes do experience same/similiar biological process.

I am a computer scientist (with one course of statistics so I have an idea) - detailed howto would be appreciated. However, advice about which method to use may be fair enough :-)

enter image description here

share|improve this question
1  
Hi whuber, thanks for your comment. x-axis represents certain positions at chromosome, y-axis frequency of some biological phenomenom. Red line is average frequency for whole chromosome. I want to show that freguency at my certain positions is significantly higher than average. (so that the biological process I observe is really there) – Perlnika Jul 10 '12 at 13:04
1  
Will try to do my best-red line is average for the whole chromosome, which consist of milions of bases, which can be understood as positions. So that we have milions of positions total. I pick up 100 positions and determine frequency of certain biological process. Because picture shows just these positions,red line does not represent average of pic. My data are from database,so my output is what I calculated using this database-there are no more measurements or data of this kind. Because I have 100 positions, data are isolated points, which were connected for better illustration of phenomenon. – Perlnika Jul 10 '12 at 13:31
1  
Good! Just a few more followup questions. (1) What are these "point" frequencies? Are they counts or are they proportions? (2) What is the basis for selecting these particular 100 points: are they randomly selected? Were they chosen because some of them exhibited high frequencies? Something else? (3) Is there any relationship to be expected among the frequencies for adjacent points in the graph? (There appears to be strong serial correlation, as evidenced by the smoothness of the line.) (4) Does the sequence of points have any meaning? – whuber♦ Jul 10 '12 at 13:31
1  
Thank you a lot! (1) counts (2) something else, they were chosen because they lie next to the elements on chromosomes called transposons, so that the positions were chosen using certain rule (3) adjacent points are neighbours on the chromosome, so yes, one could expect that two neighbours on the chromosomes do experience same/similiar biological process (4) not really, we are just showing that next to transposons there exists certain biological process, thats all what we want to show – Perlnika Jul 10 '12 at 13:37
1  
It might make sense to restructure your analysis so it analyzes data from adjacent pairs of points on the chromosome: the positions you chose by your rule and the positions near those points. Your null hypothesis would then reflect no difference between the pairs. (Nice questions by whuber!) – Joel W. Jul 10 '12 at 14:39
show 5 more comments

1 Answer

up vote -1 down vote

Based on experience with timeline (from one other diploma and my bachelor thesis) analysis I would suggest comparing coefficients of linear regression and test for autocorrelation, maybe check for breakpoints.

share|improve this answer
2  
What would be regressed against what? How would tests of autocorrelation and breakpoints answer the question? – whuber♦ Jul 11 '12 at 22:30

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.