Reply With QuoteWhat are the actual data (n1 and n2, mean 1 and mean 2,
sd 1), and why do you feel that you need to perform a test
of significance ?
Kind regards
K.
Is it possible to test a hypothesis to show that there is no significant difference between two averages without a standard deviation for both sample sets?
I have a problem involving two sets of data, salaries for location 1 and for location 2. Problem is I don't have any info on location 1 except the average, no knowledge of the std dev or sample size. I do however have this information for location 2.
Any ideas on how to show no significant difference between the two averages with such limited information?
What are the actual data (n1 and n2, mean 1 and mean 2,
sd 1), and why do you feel that you need to perform a test
of significance ?
Kind regards
K.
Is the mean for location 1 the true population mean, or the sample mean? If it's the true population mean, you could do a hypothesis test that
If not, then we might want to assume std dev for location 1 = std dev for location 2. You see, we usually do a t-test, and that test is usually based on the groups having the same std dev.
I only have 1 sample size available.
n1=60
n2=?
mean1=31,021
mean2=30,000
stdev1-3,589
stdev2-?
Both means are of the sample, not population.
You can do a one sample t-test (I think this is what mean joe meant). You would compare your sample to some theoretical mean value (normally to some other population to see if your sample differed from it signficantly). You don't need the other population SD for this.
"Facts are stubborn things, but statistics are more pliable." Mark Twain
No it would still be a two-sample t-test since we're still estimating both means. But like Mean Joe mentions we would need to make some sort of assumption. Assuming equal variances would probably be reasonable. You wouldn't have nearly as many degrees of freedom as you would if you actually had the standard deviations for both groups though.
Do you have any idea what the sample size for the first group is? If you have that then we could actually construct something fairly legitimate.
"His programming is malfunctioning. It begins! Get your weapons, he's going to become a killbot!!!" - bryangoodrich
First I thought that utoledoboi had the population values and then it is just to compare the two values and note that 31.021 are more than 30.000. (That’s what I thought MeanJoe meant.) A completely different thing is of course if it is an important difference, if it is “significant” in the non-statistical sense.
The fact that mean 2 is exactly 30.000 makes me question the data quality. There might have been done some rounding. There might be some extra uncertainty, except for the sampling uncertainty.
When you say that both means are samples, then one can ask how was the sampling done in group 2? If you don’t know the standard deviation then maybe you don’t know “where the numbers came from”. We don’t know if it was created by independent simple random sampling. Maybe not. I think that it does not take much to make the sampling skewed so that you get a larger proportion or high-income earners.
You can say something about your own group, but it seems (to me!) that it will be a little speculative to talk about the other group.
That looks like an only small difference, if we provisonallymean1=31,021
mean2=30,000
stdev1-3,589
accept SD1 as an estimator for SD2. But why do you need
to perform a test of significance even when nearly half of
the necessary information is missing?
Kind regards
K.
Posting Permissions
|
|
