T-Values

Questions about the implementation and application of the PLS-SEM method, that are not related to the usage of the SmartPLS software.
auniauji
PLS Junior User
Posts: 7
Joined: Thu Nov 22, 2012 9:32 am
Real name and title:

Post by auniauji »

Dear all,

Can we say that the study will give more impact if we use t-values of two-tailed even though our hypotheses are rather one-tail tests?

Can we, for example; the hypotheses of our study is directional and therefore we should use one tail sig. value. From the results of bootstrapping it shows that the t-values are non-significance if we use the cut-off of two tailed but accepted if one-tailed values are used. Therefore, can we just proceed using t-values for one-tailed?

Thanks.
Raha
PJ Rosenberger
PLS Junior User
Posts: 5
Joined: Fri Jun 26, 2009 1:49 am
Real name and title:

Use of one-tailed or two-tailed tests

Post by PJ Rosenberger »

Hi,

You should use one- or two-tailed tests depending on if your hypothesis states the direction of the relationship (i.e. positive/negative, use one-tailed) or does not state a direction (i.e. just that there is a relationship, use two-tailed).
Dr PJ Rosenberger III
Lecturer in Marketing

Marketing Group
University of Newcastle
Business Offices
10 Chittaway Road
Ourimbah NSW 2258
Australia
kimarthur
PLS Junior User
Posts: 2
Joined: Sun Mar 13, 2022 12:35 am
Real name and title: Kim Arthur

Re: T-Values, one-tailed or two-tailed, bootstrap

Post by kimarthur »

Hi all,


Thank you Dr Christian Nitzl for providing the following t-values:

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'm new to research and have the following questions:

Question 1: I have noticed a number of researchers in their Master/PhD theses had one-tailed hypotheses (e.g.: A is positively related to B. "OR" A has positive influence on B.). And the significance level in their theses was often set at 0.05. However, when running hypothesis testing Bootstrapping, the two-tailed test type was used, which gives them t-value 1.96, instead of 1.645 if one-tail test type is used. Why is this so?

Question 2: I have also noticed the default test type for bootstrapping in SmartPLS is two-tailed. Should I follow the default two-tailed, like everyone else, or should I change to one-tailed test which perhaps suits my study better? My study involves only one-tailed hypotheses and sets significance level at 5%.

Question 3: For running PLS Algorithm, should I select Factor or Path weighting scheme?


I greatly appreciate your help to demystify my query.
Kim
Post Reply