## standard error???

### standard error???

How do you do the standard error statistical test? I am confused. Help please

boohoo

boz2
Porifera

Posts: 3
Joined: Thu Feb 02, 2006 1:06 pm
Location: behind you

### Standard error

SE = SD/ Square root of N-1
N is the sample size
SD is standard deviation

Work out the mean, then work out the SD, then move on to standard error.

Plot a graph showing the mean for each value.

You need to multiply your SE figures by 1.96 (or 2 if you round up) then draw vertical lines up and down from the mean value. 1.96 is the SE equivalent of 95% confidence limits.

Hope this helps.

p1ckle
Porifera

Posts: 10
Joined: Wed Sep 07, 2005 7:48 pm

hey can i ask why do you have to times your mean values by 1.96 .........what does that do and what do the confidence limits mean ...............if you could help it would be kindly appreciated! thanks
B4me
Cnidaria

Posts: 16
Joined: Thu Feb 09, 2006 11:21 pm

### silly question

Does anyone know how many repeats of each sugar I would need to do to calculate standard error?
I've read somewhere that the sample size needs to be at least 30?! Does this mean 30 repeats of each sugar?

p1ckle
Porifera

Posts: 10
Joined: Wed Sep 07, 2005 7:48 pm

yes you have to have 30 or more you can pool , your results however the same experiemntal procedure has to be carried out.........if you cant gain 30 results the t test only needs 6 minimum
B4me
Cnidaria

Posts: 16
Joined: Thu Feb 09, 2006 11:21 pm

CLARIFICATION OF EARLIER WRONG POSTS (by others!)

Standard error can be used with any number of repeats - the error bars are simply bigger with smaller sample sizes.
30+ repeats are SUGGESTED for 'standard' stats tests - chi-squared, Spearmans rank. Smaller sample sizes simply mean the chances of getting a result that is stat sig are less (ie you will have to accept the null hypothesis, even if the results LOOK like there is a difference)

t-test is for comparing means of POPULATIONS (heights of males v females; lung capacity of smokers v non-smokers etc) and so has no bearing whatsoever in this context.

Chordata

Posts: 1169
Joined: Thu Oct 02, 2003 10:53 am
Location: Shrewsbury

You need to multiply your SE figures by 1.96 (or 2 if you round up) then draw vertical lines up and down from the mean value. 1.96 is the SE equivalent of 95% confidence limits.

Confidence limits means that 95% of the time any future results you obtain will lie within that range.

1.96 = 2 standard deviations

Chordata

Posts: 1169
Joined: Thu Oct 02, 2003 10:53 am
Location: Shrewsbury

Thanks every1 for the info, really helpful. keep up the good work

boz2
Porifera

Posts: 3
Joined: Thu Feb 02, 2006 1:06 pm
Location: behind you

### confused myself

I've just read somewhere that when calculating confidence limits you multiply SE by the 95% confidence value in a t table for the number of repeats in the experiment.
Is this correct?

In otherwords if I have tested 5 sugars each 20 times and the SE for one is 2.39 I would multiply this by 2.09 (read from the table DF=19) and then plot bars of this size?

I thought I'd got this but now not sure how to explain it all.

p1ckle
Porifera

Posts: 10
Joined: Wed Sep 07, 2005 7:48 pm

No - the standard error calculation takes into account the sample size.
Just plot the SE bars directly!

Added to which, the DoF iin your case is not 19, but infinite, since the size of any one result is completely independent of any other readings - they are not linked in any way. Hence, since multiplying by infinity is rather difficult, the exercise would then be rendered futile.

Hope that helps!

Chordata

Posts: 1169
Joined: Thu Oct 02, 2003 10:53 am
Location: Shrewsbury

### confidence limits

So i don't have to calculate confidence limits at all? Just use my figure for standard error.

That's great news.

p1ckle
Porifera

Posts: 10
Joined: Wed Sep 07, 2005 7:48 pm

Nope1 AQA requir SE as minimum level of stats.
Confidence limits relates to magin of error in each stage of procedure (they are cumulative), thus giving overall margin of error for the expt.
thus, measuring volume of gas to +/- 10% is not good if you are measuring volumes to 1.0%.

Hope that helps!
HD2

Chordata

Posts: 1169
Joined: Thu Oct 02, 2003 10:53 am
Location: Shrewsbury

### /N or divide by n-1

Hello. I am doing standard error and here it says to divide the standard deviation by root n-1, but elsewhere it says to divide just by n. Which is correct? Thanks in advance.
cafb
Porifera

Posts: 4
Joined: Mon Apr 24, 2006 8:55 pm

Standard error = standard deviation / square root of sample size.

Any other formulae are wrong and are calculating something else - possibly used when taking a BIG sample from a small population - possible in psychology expts at A-level, I suppose.
That won't ever happen in biology expts, though!

Chordata

Posts: 1169
Joined: Thu Oct 02, 2003 10:53 am
Location: Shrewsbury

### Error bars .. do they have to go on a bar chart!?

So do you have to plot the error bars onto a bar chart? ive plotted mine onto scatter graphs. I have a graph for each individual sugar. This means i cant see if they overlap, have i done it wrong!! As you can tell im very very confused and any help would be very appriciated!!
fuzzyduck
Porifera

Posts: 8
Joined: Sun Mar 11, 2007 8:27 pm

Next