topic Re: Steps in proving significance difference level in Statistical Procedures

Steps in proving significance difference level

koksiang100 — Sat, 08 Nov 2014 04:02:45 GMT

I have a set of data of 52 weeks of actual output and demand. In order to prove their significant difference level, can anyone please show me the step in proving the sign. level?

Re: Steps in proving significance difference level

Ksharp — Sat, 08 Nov 2014 09:17:58 GMT

proc ttest for bi-variable normal distribution data.

proc npar1way for any distribution data.

Re: Steps in proving significance difference level

koksiang100 — Sat, 08 Nov 2014 10:28:32 GMT

If I use chi square test to prove if there is any association between 2 variables? Is that ok?

Re: Steps in proving significance difference level

koksiang100 — Sat, 08 Nov 2014 10:44:58 GMT

The chi-square statistic is 12.9102. The P-Value is 0.004835. The result is significant at p < 0.05.

Re: Steps in proving significance difference level

Ksharp — Sat, 08 Nov 2014 11:33:14 GMT

That means they are associated . But you should know chi-square test is for frequency analysis. and If I don't get wrong your contingency table have lots of cells which freq less than 5 ,i.e. you should use Fisher exactly test.

Re: Steps in proving significance difference level

koksiang100 — Sat, 08 Nov 2014 11:38:19 GMT

thanks for correcting me. I wil use the approach u mentioned and give you the answer asap... thanks for the guide~

Re: Steps in proving significance difference level

Ksharp — Sat, 08 Nov 2014 11:44:32 GMT

You should know what goal you are trying to retrieve . after that , choose the right statistical method .

Re: Steps in proving significance difference level

koksiang100 — Sat, 08 Nov 2014 12:09:00 GMT

for the fisher exact test, the test is only for 2x2 table?

Re: Steps in proving significance difference level

Ksharp — Sat, 08 Nov 2014 12:33:24 GMT

no. you just need EXACT fisher statement to require this test.

Re: Steps in proving significance difference level

koksiang100 — Sun, 09 Nov 2014 01:08:09 GMT

The results of an exact contingency table test performed at 19:07 on 8-NOV-2014

data: contingency table  A      B      C      D  1     10     19     17      6     52 2      1     16     18     17     52  11     35     35     23    104    expected: contingency table  A          B          C          D  1    5.50       17.5       17.5       11.5 2    5.50       17.5       17.5       11.5

You have no "small" cells in the expected table...Use X²!

The given table has probability 1.3E-05
The sum of the probabilities of "unusual" tables, p = 0.004

Re: Steps in proving significance difference level

koksiang100 — Sun, 09 Nov 2014 01:10:02 GMT

The results of an exact contingency table test performed at 19:07 on 8-NOV-2014

data: contingency table A B C D 1 10 19 17 6 52

2 1 16 18 17 52

11 35 35 23 104

expected: contingency table A B C D

1 5.50 17.5 17.5 11.5

2 5.50 17.5 17.5 11.5

You have no "small" cells in the expected table...Use X²!

The given table has probability 1.3E-05
The sum of the probabilities of "unusual" tables, p = 0.004

Re: Steps in proving significance difference level

Ksharp — Sun, 09 Nov 2014 12:39:26 GMT

Curious. Is your contingency table one-way or two-way ?

Re: Steps in proving significance difference level

koksiang100 — Mon, 10 Nov 2014 10:36:02 GMT

http://www.quantitativeskills.com/sisa/dataprocs/table2xn.php?ts=52&inp=i m using this to help in my calculation...

btw, can I use mann-whitney test in proving the significant different?

Re: Steps in proving significance difference level

Ksharp — Mon, 10 Nov 2014 13:23:31 GMT

You misunderstand the contingency . You should firstly know what is row variable,what is column variable.

follow your mind, WEEK is row variable?

about mann-whitney test, if I am right it is almost like what I suggest proc npar1way + wilcoxon before .

Re: Steps in proving significance difference level

koksiang100 — Mon, 10 Nov 2014 13:27:29 GMT

I had some difficulty in surfing internet now to get sufficient info on 7understanding the contingency.... can u kindly explain to me how to differentiate the row n column variable in this case ?

Re: Steps in proving significance difference level

koksiang100 — Mon, 10 Nov 2014 13:44:30 GMT

In my case,dependent variable is my actual output n demand.... my independant variable is d time period(52 weeks).... am i on d right path ?

Re: Steps in proving significance difference level

Ksharp — Mon, 10 Nov 2014 13:57:45 GMT

I don't know what to say. If I am right , independant variable or dependent variable is only for general linear model or mixed model . NOT for contingency table analysis . contingency table analysis is only searching the association relationship between row variable(ACTUAL) and column variable(DEMAND) . Therefore, if you want see if they are difference in population mean value, useing mann-whitney test(i.e. wilcoxon-mann-whitney test) -- proc npar1way + wilcoxon option , check its example in sas documentation about proc npar1way .

And Hope Dr Steve could give you some constructive advice .

Xia Keshan

Re: Steps in proving significance difference level

koksiang100 — Mon, 10 Nov 2014 14:08:51 GMT

Mann-Whitney U Test in SPSS | Setup, Procedure & Interpretation

I had checked some info on mann whitney test. However i hipe you could giv me some advice on the article i paste above... in the article, assumption#2 mentioned innependant should consists of 2 categories which my independant data is time period ,do not have 2 categories.... can i still able to conduct this test ?

Re: Steps in proving significance difference level

Ksharp — Mon, 10 Nov 2014 14:23:35 GMT

The point is what is your purpose ? Check if ACTUAL is statistical different with DEMAND ? OR if ACTUAL is increasing ,so DEMAND is also increasing ?

The Mann-Whitney U test you talk about is just what I am talking about (i.e. wilcoxon-mann-whitney test).

Search npar1way at support.sas.com you will got tons of examples and documentation about it .

Re: Steps in proving significance difference level

koksiang100 — Mon, 10 Nov 2014 14:29:15 GMT

My point of purpose is actual is significantly different fron demand with the proof from statistical test.

O ok,i will go to check them out...