Portmanteau test and Shapiro-Wilk W test, Interpretation help (check)

[kathy.kinga]

Hi, I am totally new to statistics -
I would only ask for help in interpretation this two test, or confirmation if i am thinking right.

Can I write that variables a,b,c,d,e are normally distributed because '' Pr>z'' > 0.05, and f is not normally distributed because ' Pr>z'' < 0.05

How about white noise?

Is it true that b,c,d,e,f are white noise (does not have autocorrelation) because ''Prob > chi2(40)'' <0,05 and a is not a white noise (have autocorelation) because ''Prob > chi2(40)'' > 0,05


Please only check and correct my thinking if needed.

Kind regards
Shapiro-Wilk W test for normal data

	W	V	z	Pr>z
a	0.98868	0.958	-0.096	0.53822
b	0.98244	1.630	1.093	0.13726
c	0.98149	1.718	1.210	0.11323
d	0.98283	1.594	1.042	0.14865
e	0.98189	1.532	0.948	0.1714
f	0.97652	2.196	1.759	0.03930

Portmanteau test for white noise

Model	Prob > chi2(40)	Portmanteau (Q) statistic
a	0.9221	28.0682
b	0.0000	162.3201
c	0.0000	162.6615
d	0.0000	192.8795
e	0.0000	182.2451
f	0.0000	126.005

[Nick Cox]

Kathy: Please study FAQ Advice for advice about our cross-posting policy (you should tell us about it) and the spelling "Stata" (pertinent to your other posts). This overlaps with with http://stats.stackexchange.com/quest...interpretation

To understand which way round things are, try test cases where you know the answer. Here's one example.

. clear

. set obs 1000
obs was 0, now 1000

. set seed 2803

. gen gauss = rnormal()

. gen expgauss = exp(rnormal())

. swilk *

                   Shapiro-Wilk W test for normal data

    Variable |    Obs       W           V         z       Prob>z
-------------+--------------------------------------------------
       gauss |   1000    0.99906      0.590    -1.306    0.90426
    expgauss |   1000    0.57407    268.628    13.851    0.00000

. qnorm g

. qnorm e

For what's it worth, many experienced users tend to fight shy of normality tests but draw plots routinely.