JMP Formula Error While Creating Bootstrap‑Simulated Column

[#1]

I am working with a dataset that contains many missing values and a small number of observations, which makes modelling difficult. To address this, I would like to use bootstrap resampling. One column in the dataset contains the values shown in the attached Excel file, and I want to replace this column with simulated data consisting of 10,000 observations (Column Name: Sim_Y1_Liability). I used the following two formulas to generate the simulated values, but JMP returns an error when I enter them. Could anyone advise what might be causing the error and how to correct it?

JMP Formula 1:

Pick Random(

0,

2239500,

929880,

2192100,

2806020,

0,

3387600,

109080,

1057860,

441840,

880950,

1753830,

0,

968070,

1990950,

295410,

1083420,

0,

235770,

616500,

0,

279120,

0,

591090,

35100

)

JMP Formula 2:

Random Sample(

{0, 2239500, 929880, 2192100, 2806020, 0, 3387600, 109080, 1057860, 441840,

880950, 1753830, 0, 968070, 1990950, 295410, 1083420, 0, 235770, 616500,

0, 279120, 0, 591090, 35100}

)[1]

[#2]

Neither of those functions exist in JMP. You can find the available functions from Random in Scripting Index or JSL Syntax Reference JSL Syntax Reference > JSL Functions, Operators, and Messages > Random Functions . You can most likely combine your list of values with Random Integer to get one of those randomly Random Integer (jsl.jmp.com) 

[#3]

Thank you for the comments. Yes, I can generate random numbers using the technique you suggested. However, I am specifically interested in creating a bootstrap‑simulated column with 10,000 observations for the dataset. I know I can do this by selecting the distribution, clicking Mean, and then applying bootstrap. Is there an alternative method available?

[#4]

Try this JSL and see if it gives you what you need:

Names Default To Here( 1 );
dt = data table("Sample Data File_Sheet1");

dt << Distribution(
	Continuous Distribution(
		Column( :CPF_Payout_USD ),
		Process Capability( 0 ),
		Fit Normal
	)
);
obj = current report();
obj["Fitted Normal Distribution",Number Col Box( 1 )] << 
Bootstrap(
	10000,
	Fractional Weights( 1 ),
	Split Selected Column( 1 )
);

[#5]

Hi @Madhu_Acharyya,

Welcome in the Community !
Completely different answer (and questions !) from my side :

• You can have a reliable model with a small number of observations. The quality and information contained in your dataset is more important than the number/quantity of data.
• Using boostrapping to "artificially" expand the size of your dataset won't help you fit a model more easily. It will only repeat and enforce patterns already present in the data and give you a (false) impression of confidence in the model, with inflated model/performance metrics and "artificially reduced" confidence intervals.
• Based on the files you provided, it's not clear what you intend to do. I would have naively guessed that you want to do a forecasting of CPF response based on the year (so Time Series Analysis and/or Time Series Forecast) but your second file about simulations show very different columns. Can you clarify ?

If you intend to do Time Series analysis, I would try easy options first, like simple moving average with 3 or 5 points for the window width, as this simple model type still helps to understand and forecast the trend in your data:

More advanced method and models, like ARIMA / Seasoned ARIMA / ... are also available, but require more assumptions and a better definition of the seasonality and other model parameters.

Hope this answer may help in the discussion,