[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math

View post   
File: 63 KB, 353x500, 1270469838345.jpg [View same] [iqdb] [saucenao] [google]
1977287 No.1977287 [Reply] [Original]

/sci/, i have a statistics question.

I have several mean values with corresponding standard deviations. I need the mean value and the standard deviation of those values.

Could you help me out with this?

Thank you!

(inb4 highschool fag, I'm a biology PhD student and my statistics lessons were 6 years ago, inb4 OP can't inb4, I don't give a shit.)

>> No.1977296

protip: consult someone at the university you work at

>> No.1977303

I need it by tomorrow. So if anyone knows a website which explains it, I'd be happy.

>> No.1977305

need weight of each mean

else can't do it

>> No.1977315

is it possible to assume the weight of the means is equal? i have only a sample size of 2 giving each mean (don't ask, this is biology).

>> No.1977351

>>1977287
look. this is a very abstract problem; is there any real data behind those value? you generally do not do what your asking.

the mean values can be averaged, no problems if you know their weights.

however i have never encountered what your asking with the standard deviation. you don't ever do that.

logically, what your doing is putting many data sets together, and then pulling out a value and asking about its properties..

>> No.1977375

http://en.wikipedia.org/wiki/Sum_of_normally_distributed_random_variables

this is closest thing, the sum, but it's not exactly what your asking.

>> No.1977430

>>1977375
>>1977351

OP here, I know very little about statistics (bad thing about our university, very little education in statistics). If you say you don't do that, I believe you. My problem is as follows:

- Several experiments yielding a value of a fluorescence intensity
- I have 2 untreated (control) samples and 2 treated samples (so n=2) per experiment
- I average the 2 controls and the treated samples and calculate the relative values and corresponding standard deviations (control is then 1, treated sample is x times control)
- i have several of those experiments and want to find a mean value with a standard deviation which "summarizes" all experiments.

Thanks anyway for you opinion and the link, i will have a look at it.

>> No.1977442

>>1977430

or let it be an error bar instead of a SD...

>> No.1977470

>>1977430
i suggest you accumulate all the original data and then find what you need.

whatever manipulation you do to standard deviations, the result should converge to the 'true' standard deviation (which you can simply get if you accumulate all original data)

sorry about my english -- not so good

>> No.1977554

>>1977470

OP here, this is what i initially thought of, but I was looking for a better way to do it. Thank you!

>> No.1977588

>>1977287
>PhD student and my statistics lessons were 6 years ago

BOOKS!!! You know them??? They have all those silly little signs inside them that make some sort of sense if you combine them while looking at them!

>> No.1977894

Keywords 'propagation of error':

http://en.wikipedia.org/wiki/Propagation_of_uncertainty

Don't give up, there's a reasonably simple table of examples in the middle.

You WILL need to go back to the data set to find the covariance (unless there's you'd dead certain it's random).

It's so bizarre that you got ten people spouting random answers. I guess /sci spends a lot of time thinking about how smart it is, and not a lot of time conducting experiments...