what do SEM+,- and *p<0.05 mean? - (Jun/05/2009 )

hi all

I don't know what do the following terms mean? can somebody please give me some info?

I have 4 separate sets of samples and I have calculated the means of all. I need to draw bar graphs with error bars. but why do I need to add SD or SEM? and how to calculate p? why put * ?

I'm lost and confused

thanks.

SEM is Standard Error of the Mean.
SD is Standard Deviation. These terms are measures of the scatter about the mean and really should be used with normally distributed data and with a sufficiently sized sample group.
p<0.05 is the probability that the results are NOT different (the null hypothesis). In this case, the probability is 95%.

A mean needs to be qualified to show statistical variation. This may cover biological variation as well as analytical variation. Try pipetting 450ul twenty times and weighing each drop. Same thing, try shooting at a target with a rifle ten times. You wont get exactly the same weight of drop and you wont get each bullet to go through the same hole. This is called imprecision and is partly a characteristic of the rifle or pipette and partly of you (more so in the case of the shooting). Your breath volume differs slighlty, muscle fatigue, eye relief etc all play a part. Inside the pipette, springs and other moving parts have tolerances of function.

Its all to do with how confident you are that the next 450ul drop will weigh exactly the same as the mean or that the next bullet will go through the same hole as before.

So you need to state the mean weight of the 450ul drops together with the range. But what range? Heaviest and lightest? Well, conventionally we use SDs or SEMs. There is a rigourous mathematical reason for this and you should check a statistics text book for that.

When you are reviewing data e.g. in a paper, looking at the SEMs and Sds can help you spot unlikely data...made up?

Hope that helps.

In addition to what has been posted, in case you are unsure of how to get SD and SEM, if you are using Excel it will do it for you. You can go into the function menu and find SD or you can type in an empty box =STDEV(highlight your means here) and it will find SD. We use this number to get SEM, which is SD/Sqrt(n) where n is the number of means you have. We use this for graphing. Your means or averages are your graphed numbers, then the SEM you can add to each average as your error bars.

Example: Your numbers are 5.3, 4.5, 4.8, 5.1 Ave=4.925 Stdev=0.35 SEM=0.35/(SQRT(4)) Which =.175
SO you graph your average 4.925, and add 0.175 for your error bar. If you error is very high it is likely your differences between this group and another group may not be significant and need statistics to verify.

Curtis on Jun 5 2009, 05:45 AM said:

thank you very much guys,

your explanations and examples were really helpful.

just one quick question which goes to labmeiser; in the above example, what would be the p value?