Chi Squared

[Of course]

1000 random numbers in 20 categories. How many degrres of freedom you guys used? I wasn't sure between 19 and 20 but i picked 20...

[Strategery]

Originally Posted by Of course

1000 random numbers in 20 categories. How many degrres of freedom you guys used? I wasn't sure between 19 and 20 but i picked 20...

19 I think. (# groups) - (# estimated parameters) - (1). You didn't estimate any parameters, so it's just 20-1 = 19. I don't believe it would be 20 under any circumstances. Someone call me out if I'm totally off on this.

[jadr]

I think it was E for the Chi-Squared, was that right? I just remember getting a statistic that was way larger than the values of that in the Chi-Square distribution table so you reject at all levels?

[Strategery]

Yes, now that you mention it I think I got E too. I know there was one Chi-Square problem where I rejected at all levels, so it might have been that one.

[ultfrisbee]

37 was my stat

[green_alien]

Originally Posted by Strategery

Yes, now that you mention it I think I got E too. I know there was one Chi-Square problem where I rejected at all levels, so it might have been that one.

Is this the likelihood ratio test problem? I recall rejecting it all levels.

[Strategery]

Originally Posted by green_alien

Is this the likelihood ratio test problem? I recall rejecting it all levels.

Hmm, now I'm confused. For the likelihood ratio test, I'm almost certain I rejected at all levels....was there only 1 degree of freedom for that? That's what I assumed, and I rejected everywhere.

So maybe I rected at all levels for both of these problems, but for sure the likelihood ratio test. I was less confident in that answer though.

[green_alien]

Originally Posted by Strategery

Hmm, now I'm confused. For the likelihood ratio test, I'm almost certain I rejected at all levels....was there only 1 degree of freedom for that? That's what I assumed, and I rejected everywhere.

So maybe I rected at all levels for both of these problems, but for sure the likelihood ratio test. I was less confident in that answer though.

I assumed that df = 1, as well and got a large answer that it had to be rejected at all levels. I guessed on the pseudo-U(0,1) problem, though.

[mreevit]

Originally Posted by Of course

1000 random numbers in 20 categories. How many degrres of freedom you guys used? I wasn't sure between 19 and 20 but i picked 20...

How the hell was this one done?

you're given 20 intervals with 1/20 * 1000 expected in each one. So you need the sum of (O-E)^2/E for 20 observations. the expected for each interval was the same so ...the test stat is

(Sum(O^2) -SUM(100*O) +20*20^2)/20


all of this can be figured out but the Sum(O)

How did you calculate this?

[beck]

Originally Posted by mreevit

How the hell was this one done?

you're given 20 intervals with 1/20 * 1000 expected in each one. So you need the sum of (O-E)^2/E for 20 observations. the expected for each interval was the same so ...the test stat is

(Sum(O^2) -SUM(100*O) +20*20^2)/20


all of this can be figured out but the Sum(O)

How did you calculate this?

i was wondering the same thing, the answer seemed to be JUST there, but i couldnt get it .....