T-Values

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
seb
PLS Junior User
Posts: 1
Joined: Tue Oct 10, 2006 11:32 am
Real name and title:

T-Values

Post by seb »

I know that the T-Values have to be over 1,65 to be valid. Can anybody tell me about the minimum T-Values for the Path Coefficients?
User avatar
Diogenes
PLS Super-Expert
Posts: 899
Joined: Sat Oct 15, 2005 5:13 pm
Real name and title:
Location: São Paulo - BRAZIL
Contact:

Post by Diogenes »

Hi,

when we use t-values, and say that t-values are significant (t>1,65 for example), We are rejecting the hipotesis that correlation or path coefficient is equal to zero in the population (it´s significant).

Then you choose a significance level (90%, 95%, 99%, 99,9%) to have the critical t-value (minimum to be significant), or compute the significance (p-value) from the t-value presented in SmartPLS output.

Best regards

Bido
viswadatta
PLS Expert User
Posts: 139
Joined: Wed Jul 05, 2006 1:43 pm
Real name and title:
Location: Coimbatore, India

t-values

Post by viswadatta »

Use the following values
Two Tail
90% significance level--t value-1.645
95% significance level--t value-1.96
99% significance level--t value-2.57

One tail
90% significance level--t value-1.628
95% significance level--t value-1.645
99% significance level--t value-21.96
viswadatta
PLS Expert User
Posts: 139
Joined: Wed Jul 05, 2006 1:43 pm
Real name and title:
Location: Coimbatore, India

t-value

Post by viswadatta »

Sorry for one tail with 90% significance value the t-value is 1.28 not 1.628
droople
PLS Junior User
Posts: 6
Joined: Wed Aug 15, 2007 2:05 pm
Real name and title:

Post by droople »

Hi

I supposed that the t-statistic is actually z-statisitic as the values are the same as the Z table, am I right?

And should I use two tails or one tail?

Thanks
viswadatta
PLS Expert User
Posts: 139
Joined: Wed Jul 05, 2006 1:43 pm
Real name and title:
Location: Coimbatore, India

number of tails

Post by viswadatta »

Since the bootstrap sizes are 100 or more, the values are actually z-values.

By default two tail values are better, although one tail values need smaller t-values to prove significance
droople
PLS Junior User
Posts: 6
Joined: Wed Aug 15, 2007 2:05 pm
Real name and title:

Re: number of tails

Post by droople »

viswadatta wrote:Since the bootstrap sizes are 100 or more, the values are actually z-values.

By default two tail values are better, although one tail values need smaller t-values to prove significance
Thanks a lot, I found literature said if the hypotheses is directional, one tail should use.
fabio.colacchio
PLS Junior User
Posts: 7
Joined: Thu Mar 22, 2007 4:47 pm
Real name and title:

Post by fabio.colacchio »

Can you give me an example of what is "one tail" and what is "two tails"?
thanx
Huet
PLS Junior User
Posts: 4
Joined: Thu Jul 28, 2011 1:32 pm
Real name and title:

Post by Huet »

I want to check my formative constructs (latent variables) in my model.

I know I have to check if the path coefficients from the indicators to the construct are significant. After bootstrapping, I got the t-values.

What should I do now?

I already calculated the p-values for each t-value. Have done it with the calculator from this site: http://danielsoper.com/statcalc3/calc.aspx?id=8

Example:
for the t-value of 2.525 the calculator says

Probability (one-tailed): 0.00586183
Probability (two-tailed): 0.01172366

How can I interpret this p-values? If I choose the two-tailed one, can I say its significant with a probability of error of 5% (because p<0,05), if I choosed my levels before as p<0,05, p<0,01 and p<0,001?

Another thing is, I am confused about if I should use a one-tailed t-test or the two-tailed t-test. In which case I have to use the one or the two tailed test?

And will be the whole procedere nearly the same if I wanna check the path coefficients of my inner model?

Please help!
Thanks!
christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl »

We typical use the two-tailed test to check the weights and also the path coefficients.
cn
firm4n2003
PLS Junior User
Posts: 1
Joined: Thu Aug 04, 2011 2:21 am
Real name and title:

Post by firm4n2003 »

Do we need to determine the degree of freedom before deciding the t-value?, for example i have 101 observations, how do i determine the t-value significant level?
iRose
PLS User
Posts: 13
Joined: Sat Mar 05, 2011 2:36 am
Real name and title:

Re: t-values

Post by iRose »

viswadatta wrote:Use the following values

One tail
90% significance level--t value-1.628
95% significance level--t value-1.645
99% significance level--t value-21.96
I think there may be a typing error for the t-value (one-tail) for 99% significance level. It is 2.326 and not 21.96. Please correct me if I'm wrong.

Thanks!
christian.nitzl
PLS Expert User
Posts: 248
Joined: Sat Jul 25, 2009 1:34 pm
Real name and title:

Post by christian.nitzl »

Sorry, there are more typing errors than one. Firstly, we typical use a two-tailed test. The critical t-values for a two-tail test are:

90% significance t-value = 1.64
95% significance t-value = 1.96
99% significance t-value = 2.58

For a one-tailed test the critical t-values are:

90% significance t-value = 1.28
95% significance t-value = 1.64
99% significance t-value = 1.96

I hope that helps!
Christian
karenkim
PLS Junior User
Posts: 6
Joined: Sat Jan 07, 2012 5:06 pm
Real name and title:

Re: number of tails

Post by karenkim »

droople wrote:
viswadatta wrote:
I found literature said if the hypotheses is directional, one tail should use.
Can you tell the literature you found?
Mrs Tourky
PLS User
Posts: 11
Joined: Sat Dec 03, 2011 3:31 pm
Real name and title:

Post by Mrs Tourky »

Hi Namhyun,

You can find this in any statistical reference/textbook, not necessarily PLS reference.

Marwa
Post Reply