# What test to use?

#### Anous

##### New Member
Hello
I hope everyone is doing well. I have the mean scores of the dependent variables. I want to see whether their order is sgnificant. What statistical test to use? Attached is a file with the data.

#### Attachments

• 204.9 KB Views: 2

#### obh

##### Active Member
Do you have any assumptions on the order, or is the order is only the sample results?

You may use multiple comparisons with significance correction, or tukey hsd test.

You may just calculate a confidence interval per each and have a chart

#### Anous

##### New Member
The research question is: How does grammatical complexity affect the acquisition order of conditional types?
Attached are the mean scores of the diffrent conditional types.
How can I use the significance correction?
How can I calculate a confidence interval per each?

#### Attachments

• 204.9 KB Views: 1

#### obh

##### Active Member
Hi Anous,

Is it research or assignment? say do you have the data or only the averages/standard deviations?
Are SD known SD? or sample SD?

Do you have to do it with SPSS?

If you have the data, you can use one-way Anova with Tukey HSD :
The Anova will answer the default assumption that all averages are equal, and the Tukey the order:
http://www.statskingdom.com/180Anova1way.html, includes Tukey HSD, You will also get the R code so you could then run on R if you prefer.
If you use Tukey HSD you don't need to do correction,

Alternatively, you may run in R the pairwise.t.test with correction.
The correction options build in the function
I would probably use the Holm

Confidence interval:
http://www.statskingdom.com/40_confidence_interval.html

Or in R:

mean=10
conf_level=0.95
alpha=1-conf_level
s=5
n=18
moe=qt(1-alpha/2,df=n-1)*s/sqrt(n)
mean+moe
mean-moe

Similarly, if you know the standard deviation, not sample SD, use the normal distribution

#### Anous

##### New Member
Thank you very much. I have the data from which I got the means and standard deviations. Anyway, I found the useful test which is Friedman test.

#### obh

##### Active Member
Why do you prefer the non parametric test?

Does the distributions non symmetrical?

#### Anous

##### New Member
Yes. After submitting the normality test, the result shows that the data is not normally distributed

#### Anous

##### New Member
My research field is second language acquisition. What about you?

#### obh

##### Active Member
Anova and t test are not sensitive to normality, what is the sample size?

#### Anous

##### New Member
I think they are sensitive to normality. That's why I adopted Friedman test.

#### obh

##### Active Member
The test's assume normal distribution but not sensitive to deviation from the normality.

It is also depend on the sample size, what sample size did you use to calculate the averages? (clt)

#### obh

##### Active Member
120 per each average? From 30 you can use the t test or ANOVA

#### Anous

##### New Member
OK. Thank you very much. I highly appreciate your help.

#### obh

##### Active Member
Great I would probably use Anova plus Tukey HSD