Hi all
I run a nonlinear sur using "nlsur"
I then run a test on nearly 8 nonlinear restrictions on the resulting coefficients using "testnl". I add to "testnl" the option, mtest(noadjust), which is the default and then alternatively mtest(holm).
The resulting outputs are as following:
and then the Holm adjustment leads to :
My question is:
Focusing on the unadjusted p_values and the results on the test of restriction number (6). When I comment on the significance (using a 5% significance level), I say that there is a significant difference with regard to restriction number (6) as the p-value of 0.026 is less than 5%.
Now, with regard to the Holm adjusted P-value (which is really new to me but was recommended by a reviewer), do I also compare between the Stata output of Holm adjusted p values i.e. 0.107 in restriction number (6) and the same 5% significance level and hence my comment would be that there is no significant relation (because 0.107>5%) ? or that I must also adjust the 5% conventional significance level by dividing by number of tests which is 8 here??
In other words, does Stata give us the adjusted p value that we directly compare with the conventional significance levels, for example 5%, or that we are also required to adjust the significance level of 5% as well before comparing with the Holm adjusted p values and make comment on significance??
I run a nonlinear sur using "nlsur"
I then run a test on nearly 8 nonlinear restrictions on the resulting coefficients using "testnl". I add to "testnl" the option, mtest(noadjust), which is the default and then alternatively mtest(holm).
The resulting outputs are as following:
Code:
---------------------------------------
|F(df,190574) df p
-------+-------------------------------
(1) | 414.41 1 0.0000 #
(2) | 4.24 1 0.0395 #
(3) | 98.97 1 0.0000 #
(4) | 2.51 1 0.1131 #
(5) | 35.04 1 0.0000 #
(6) | 4.91 1 0.0268 #
(7) | 1.96 1 0.1616 #
(8) | 5423.86 1 0.0000 #
-------+-------------------------------
all | 1230.26 8 0.0000
---------------------------------------
# unadjusted p-values
Code:
---------------------------------------
|F(df,190574) df p
-------+-------------------------------
(1) | 414.41 1 0.0000 #
(2) | 4.24 1 0.1185 #
(3) | 98.97 1 0.0000 #
(4) | 2.51 1 0.2262 #
(5) | 35.04 1 0.0000 #
(6) | 4.91 1 0.1070 #
(7) | 1.96 1 0.1616 #
(8) | 5423.86 1 0.0000 #
-------+-------------------------------
all | 1230.26 8 0.0000
---------------------------------------
# Holm adjusted p-values
Focusing on the unadjusted p_values and the results on the test of restriction number (6). When I comment on the significance (using a 5% significance level), I say that there is a significant difference with regard to restriction number (6) as the p-value of 0.026 is less than 5%.
Now, with regard to the Holm adjusted P-value (which is really new to me but was recommended by a reviewer), do I also compare between the Stata output of Holm adjusted p values i.e. 0.107 in restriction number (6) and the same 5% significance level and hence my comment would be that there is no significant relation (because 0.107>5%) ? or that I must also adjust the 5% conventional significance level by dividing by number of tests which is 8 here??
In other words, does Stata give us the adjusted p value that we directly compare with the conventional significance levels, for example 5%, or that we are also required to adjust the significance level of 5% as well before comparing with the Holm adjusted p values and make comment on significance??
