uncommon statistics

[markman]

I have to see if some raw data is significantly different from an average with an SEM. Is that possible? is if so, what test do I use?

...not sure if you can compare raw data with an average....

[jim mcnamara]

What you can do is calculate the mean & standard deviation for the population - the sample of raw data you have.

A very simplistic analysis is calculate the new mean +/- (2* std dev).

If the old mean falls within that range, the samples are probably the from the same 'population'. ie., they are stastictically similar.
This isn't a perfect analysis, there are much more sophisticated tests, but if something falls inside the 95% confidence interval (what this tests), then nobody will question a statement of similarity.

Where you run into problems is deciding if they are statistically different.

If you want a headache go to www.mathworld.com.

Look up chi square distribution, chi-squared test, Pearson's function.

[noble]

a verrry simply solution without the need to learn
statistics and distributions would be simply load your
raw data into excel and use it's built in mean and std deviation
functions. You can have graphical representations of
the distribution of your data and can analyze it a lot easier.

Also, be careful about how you decide if two samples are
statistically similar. It is a rash decision to assume that
two distributions having similar means are statistically similar.
In this case, however, since you are only concerned about an
average value (or mean) jim is correct that you can simply
compare the means and assume similarity.

[markman]

Ive done the std and sem and stuff, but I need to compare 2 lines and see if they are significantly different. Sometimes, both the lines have error bars, and sometimes one has error bars and one doesnt. If you just know the name of the statistic needed, thats all i need

[markman]

I went to a lab near me, and they said to choose 2 points, then do a t-test on them