# Need help to test the significance

#### Lisa A

##### New Member
I want to make a test to see if the difference is significant or not. What I have is two different groups of values (one with 31 values and the other with 28 values). These values have nothing to do with each other, (since it is values of iron intake for people in two different cities). I need to see if the difference between these cities is significant or not. I think it is enough with one star significance. I have all values in excel. I would be very thanksful for an answer.
I think one way to do it is to write TTEST("vales from the first city";"values from the second city";?;?), but I do not know what should be instead of the "? marks".

#### Karabiner

##### TS Contributor
Have you already tried the help function? As far as I can remember, you have
to indicate whether it is a one-sided or two-sided test, and which type of t-test
you want to carry out (dependent versus independent samples). Check the
help function for details.

Last edited:

#### Lisa A

##### New Member
Thanks for your answer. Yes, I have checked the help function, but I don´t understand how I can figure out if it´s a one- or two-sided test... I do not understand the difference and most important, I do not know in which group my data fit.

"and which type of t-test you want to carry out (dependent versus independent samples" Yes, this is the other problem.... How do I know that? I mean, what does dependent and independent samples means and how do I know which one I should chose in this case?

#### Karabiner

##### TS Contributor
From your description I would conclude that you perform a t-test for independent
samples -- if you have measured 31 and 28 = 59 different people.

Regarding one-sided versus two-sided, choose two-sided. It tests whether
there is any statistically significant difference of means between the two
groups (cities), regardless of its direction. Again, regarding Excel that is what
I remember, maybe I've forgotten one or two important points.

Regards

K.

#### Mean Joe

##### TS Contributor
I think one way to do it is to write TTEST("vales from the first city";"values from the second city";?;?), but I do not know what should be instead of the "? marks".
For the third parameter, put 2. That will let you do a two-tailed test, ie a test to see if the iron intakes are equal in city A and city B. FYI a one-tailed test would test if one city is higher than the other, and gives a p-value equal to $$\frac{1}{2}$$ of the two-tailed test.

For the fourth parameter, you should put either a 2 or a 3. Put 2 if cities A and B have equal variance, but put 3 if they have unequal variance.

#### astro_girl

##### New Member
I used the NIST freeware stastistics package called DATAPLOT. I had a similar data set that I wanted to compare for my dissertation, and with some help from an engineering statistician there, he suggested I look at the data all possible ways.

After verifying that my data were normally distributed, I performed an ANOVA and t-tests, as well as the Levene test and f-tests. The Levene test (Table 7) was performed to test for equivalence of variation (population standard deviations) among the three temperature groups of data (different sample sizes) that I compared. The ANOVA were performed to test for shifts in location (population means) among the three temperature groups. These tests were used to compare the three groups simultaneously.

The F-test was then performed across the 3 temperature groups to obtain more specific information about the Levene Test results. The t-tests further support what's evident from the graphical analyses (that I performed, which you'll need to do first) and the ANOVA. The f- and t-tests were used for group-to-group comparisons (comparing 2 groups at a time as opposed to all 3 at the same time) ...indicating a more individualized result. Does this make sense?