30b Probability Bug? Or Feature?

[Les Wright]

Hi folks, just got a 30b which I hope to flash as a 34s once I get the cable from Gene.

In the meantime, I have been fooling with the native calculator. I just happen to be interested in probability distributions, so I fiddle with that.

In RPN mode, I try this: I want to compute the upper-tail probability associated with 2 and 3 df at F = 5. The 30b gives lower-tail probabilities, so instead I have to compute the lower-tail probability associated with 3 and 2 df at F = 1/5.

If I just put the three arguments directly on the stack (3 INPUT 2 INPUT .2 Math -> Probability -> F-Dist) a correct answer is put in the display register. (It's approximately 0.17117.)

But if I COMPUTE the final argument, it doesn't work: 3 INPUT 2 INPUT 5 1/x Math -> Probability -> F-Dist, nothing happens to the contents of the display.

I think the same problem exists with other probability distributions, but I haven't experimented extensively.

Anyone have a clue? These functions aren't terribly useful if they only accept entered and not recalled or computed arguments.

TIA,

Les

[Gene Wright]

The functions in the math menu were intended to function as an extension of the keyboard. That's why if you key a value and walk through the trig functions the display changes - it is showing you the result IF you press INPUT / = to choose the displayed function.

I don't have a 30b handy (they are all 34S units now) :-) but does it give the correct answer for the distribution if you press INPUT or = ?

If not, there could be a problem with this 3 argument function which is admittedly a tough one for this paradigm.

(and your cable should arrive any day now...I mailed 3 more to people today too).

[Mark Storkamp]

I'll answer for him since I've got my 30b handy. No it doesn't, it returns with the .2 still in x. It also doesn't seem to work when you press INPUT after typing in the .2 (the INPUT doesn't, on this model, push the .2 into Y). Otherwise mine works just as he described.

[Thomas Radtke]

The 20b, using the firmware from the SDK, returns an error. OTOH, I have no clue what I'm doing here since F-Distribution is completely unmentioned in my manual, it's not even shown in the menu scheme.

Maybe a straight RPN implementation instead of the RPN/RPL approach given would have been less prone to errors like these. Just speculating ...

[Gene Wright]

Not mentioned in the small printed guide? It should be mentioned in the PDF manual at least, and I know it is in the learning modules. ;-)

Will it work if you compute the 0.2 then enter the other arguments, rearrange the stack, etc. then go to the F-dist?

It should work as you have tried, but now it is a matter of trying to figure out under what circumstances it works vs. not.

[Thomas Radtke]

Quote:

---

Will it work if you compute the 0.2 then enter the other arguments, rearrange the stack, etc. then go to the F-dist?

---

No. But it's not a bug: The manual you've pointed me to explicitly requires the user to *type in* the value for the probability. Either it has been deliberately designed that way or the manual was done to reflect that glitch. You can probably tell which way it was, Gene :-D.

[Gene Wright]

Suspicions are not proof! lol

I am not aware of any changing of the manual because of a bug. I do agree computing the argument would be best, but ...