topic Re: how to convert p value of <.0001 to actual number in Discussions

how to convert p value of <.0001 to actual number

jaboraws — Tue, 20 Sep 2016 20:18:21 GMT

Is it possible to get an actual number for pvalues instead of "<.0001"?

Re: how to convert p value of <.0001 to actual number

ms — Tue, 20 Sep 2016 20:34:27 GMT

In report table boxes, double-click on a p-value (or any number col box) to bring up the number format window. Change the setting from "PValue" to "Best".

Re: how to convert p value of <.0001 to actual number

jaboraws — Tue, 20 Sep 2016 20:40:30 GMT

Thanks but I have already tried that and the "<.0001" values just get converted to "0".

Re: how to convert p value of <.0001 to actual number

ron_horne — Wed, 19 Oct 2016 04:06:54 GMT

try changing it to fixed Dec as follows:

Re: how to convert p value of <.0001 to actual number

jaboraws — Tue, 20 Sep 2016 20:56:19 GMT

Thanks but I also tried that, even maxing out the width and decimal places and still converts is to simply "0" or "0.00000..........".

Re: how to convert p value of <.0001 to actual number

ms — Tue, 20 Sep 2016 20:58:56 GMT

Try to edit the width of the number. But maybe there is a limit for how small numbers JMP will show.

Here's a really small p-value from an ANOVA (I use JMP 12.2).

Prob > F

4,3e-179

Re: how to convert p value of <.0001 to actual number

jaboraws — Wed, 19 Oct 2016 04:07:04 GMT

Thanks again, I tried that as well and it still just converts it to "0". Actually the p value you got is exactly what I want as well. Not sure why I cannot convert it like you. I also am using JMP12.2.

Example:

Re: how to convert p value of <.0001 to actual number

martindemel — Wed, 21 Sep 2016 14:17:44 GMT

Hi,

Have you tried to use scientific format rather than fixed Dec?

Re: how to convert p value of <.0001 to actual number

jaboraws — Wed, 21 Sep 2016 14:21:31 GMT

Yes, I have tried all the different formats and different widths and decimal places. It all just converts to "0".

Thanks though.

Re: how to convert p value of <.0001 to actual number

Steven_Moore — Wed, 21 Sep 2016 14:26:29 GMT

jaboraws,

JMP uses a value called Log Worth in the Partition platform to convert very small p-values into useable numbers. You might want to look at that.

The Log Wroth is the -log10(p-value).

Hope this helps.

Re: how to convert p value of <.0001 to actual number

martindemel — Wed, 21 Sep 2016 14:34:58 GMT

The reason why you see 0e+0 is most likely that the number of leading zeros after the decimal is higher than what you asked to display. Then JMP or any other software can only represent what is given by the calculations which is 0 then.

However, you should be aware of floating point arithmetic and what it is capable of calculating.1e-17 usually is the point where a floating point number cannot differentiate from 0, which means every non-zero value smaller than 1e-17 will be interpreted in calcualtions as zeros. That's (within others) for what symbolic math is good for, because you use symbolic values to deal with that, and the maths behind will calculate further on with mathematical rules rather than actual numeric calculations. Just a side note

Re: how to convert p value of <.0001 to actual number

martindemel — Wed, 21 Sep 2016 14:37:03 GMT

Hi MS, could you provide the example for this ANOVA value, I'm just curious how much I can force JMP to show those values Thanks!

Re: how to convert p value of <.0001 to actual number

jaboraws — Wed, 21 Sep 2016 14:49:16 GMT

This is exactly what I am trying to do. I performed a Fit Y by X and then performed a Dunn's test to compare the means of my samples versus my control group and got a list of p values (which I ultimately want to convert to -log10(pvalues)).

The problem is the list of p values contains many "<.0001" values which are not usable for calculating the log worth. I am trying to look into the log worth calculation in the partition but not having much luck yet.,

Thank you for the information.,

Re: how to convert p value of <.0001 to actual number

martindemel — Wed, 21 Sep 2016 15:20:52 GMT

Question, can you share the data privately, so I could take an explicit look on it? Guess it is at least too confidential to share it here, but if you want me to try to replicate what you experience, I'm happy to take a look.

just email to martin.demel@jmp.com

... and the steps you did

Re: how to convert p value of <.0001 to actual number

ms — Tue, 15 Sep 2020 21:35:56 GMT

Here's an example. If I push it a little further, e.g. by decreasing the std dev, the ANOVA p-value will show as "0". For Dunnett's method, low p-values are less accurate. They appear to show as "0" if lower than about 1e-15.

dt = New Table("test", add rows(2e3), New Column("X", character), New Column("Y", numeric));
For Each Row(
    If(Row() > 1e3,
        :X = "B";
        :Y = Random Normal(0, 50);
    ,
        :X = "A";
        :Y = Random Normal(1e2, 50);
    )
);

dt << Oneway(
    Y(:Y),
    X(:X),
    With Control(1, {"A"}),
    Means(1),
    SendToReport(
        Dispatch({"Oneway Anova", "Analysis of Variance"}, "Prob > F", NumberColBox, {Set Format("Best", 12)}),
        Dispatch(
            {"Means Comparisons", "Comparisons with a control using Dunnett's Method", "LSD Threshold Matrix"},
            "p-Value",
            NumberColBox,
            {Set Format("Best", 12)}
        )
    )

);

Prob > F

6,90185e-311

p-Value

Re: how to convert p value of <.0001 to actual number

rlw268 — Tue, 15 Sep 2020 20:56:28 GMT

Did you ever find an answer for this? I am using logistic regression one of my x variables has a p-value of 0.00001 and logworth 56.125, and the other has a p-value of 0.00219 and logworh of 2.something. I know JMP is taking the negative log of the p value but that does not work for p-values of 0. I am lost?

Re: how to convert p value of <.0001 to actual number

Mark_Bailey — Wed, 16 Sep 2020 13:40:16 GMT

You can double-click the p-value, change the format to Best, and change the Width so that you can see the whole number, for example, Width = 15.