Challenge:Help with t-test using only averages, standard deviations and sample sizes!

#1
I want to check if the difference in my findings are statistically significantly different from the findings of an earlies study. The problem is that I do not have the full data-sets of the earlies study.
"
This is the data from the earlier study:
(.46 percent average) [t(15) = 5.39, SEM = .06,p < .001],
"
It also mentions that there were 48 participants.

This is my data set:

50
33,333332
66,666664
0
0
33,333332
50
50
83,333336
0
33,333332
100

- Gives the average of 41,6666663333

Now is my average of 41,6 % statistically significantly different from the 46 % found by the earlies study? Help is very appreciated!

BTW: I only need to know how to do this in a computer program like gnumeric, open office calc or similar...! I need the results not a huge explanation how to do it manually.
 
Last edited:

Dason

Ambassador to the humans
#2
Re: Challenge:Help with t-test using only averages, standard deviations and sample si

It looks like the percentages you have are actually results from a binomial experiment. Do you have the number of success and number of trials for each of those data points?
 
#3
Re: Challenge:Help with t-test using only averages, standard deviations and sample si

It looks like the percentages you have are actually results from a binomial experiment. Do you have the number of success and number of trials for each of those data points?
Thanks for answering.

It's not a binominal experiment as far as I understand.

Each number in my table is one individuals average percent performance in an experiment. So the end number is one groups average performance in percent. The other experiment has the identical setup, but as you can see the problem is that I don't have that experiments total data for comparison. I only have their results which I posted.

Now I got a tip that it should be possible to calculate the standard deviation from those results and then use a t-test using only averages, standard deviations and sample size, but I can't figure out how...
 

Karabiner

TS Contributor
#4
Re: Challenge:Help with t-test using only averages, standard deviations and sample si

This is the data from the earlier study:
(.46 percent average) [t(15) = 5.39, SEM = .06,p < .001],
It also mentions that there were 48 participants.
If it's a one-sample t-test, then the degrees of freedom indicate n=16.
Now is my average of 41,6 % statistically significantly different from the 46 % found by the earlies study?
It isn't. And your data don't meet the assumptions for a t-test. But anyway,
SEM (Standard Error of the Mean) = Standard Deviation / SQuareRoot of n .
If I understood you correctly, then you already know online t-test calculators
where you only need to put in the n's, means and standard deviations.

Kind regards

K.
 
#5
Re: Challenge:Help with t-test using only averages, standard deviations and sample si

If it's a one-sample t-test, then the degrees of freedom indicate n=16.

It isn't. And your data don't meet the assumptions for a t-test. But anyway,
SEM (Standard Error of the Mean) = Standard Deviation / SQuareRoot of n .
If I understood you correctly, then you already know online t-test calculators
where you only need to put in the n's, means and standard deviations.

Kind regards

K.
Oh, maybe that only included a sub-group of the participants then.

What p-value did you get when you say it's not statistical significant? - And I'm sure it's supposed to be a t-test! (confused)
Ok then I know how to get the standard deviation.

No, I only know how to do a t-test with two colums of data in the program. Not with only the averages, standard deviations and sample size. - Maybe it's not even a function in open office calc or gnumeric?
I'm at my first year on the university and do not have great insight into statistics. Appreciate the help!!
 

Karabiner

TS Contributor
#6
Re: Challenge:Help with t-test using only averages, standard deviations and sample si

What p-value did you get when you say it's not statistical significant?
None, I'm afraid. Just a quick estimation. Your 0.044 mean difference
is minimal, compared to the SD in the reference group and the SD in
your own sample obviously is large. So with a total sample size of only
28 there will surely no p < 0.05 test result.
I'm at my first year on the university and do not have great insight into statistics.
You do not need that much insight. Some basic search skills
will do for the moment http://tinyurl.com/6td83d4 .

Kind regards

K.
 
#7
Re: Challenge:Help with t-test using only averages, standard deviations and sample si

None, I'm afraid. Just a quick estimation. Your 0.044 mean difference
is minimal, compared to the SD in the reference group and the SD in
your own sample obviously is large. So with a total sample size of only
28 there will surely no p < 0.05 test result.

You do not need that much insight. Some basic search skills
will do for the moment http://tinyurl.com/6td83d4 .

Kind regards

K.
Ok I'm going to trust you on that then! Thanks for explaining. I knew it was a long shot.

I can promise you that I did do a lot of searching before asking here. Anyways, thank you for your help...!!