Have an idea?

Visit Sawtooth Software Feedback to share your ideas on how we can improve our products.

SoP standard errors on weighted sample

Hi!

When one exports an Excel-simulator from Lighthouse Studio, there are formulas calculating standard errors for Shares of Preference.

For example, for Product 1 it stands:
=SQRT((SUMPRODUCT('Share Calculation'!T2:T288;'Share Calculation'!T2:T288;FiltersWeights!D2:D288)-(SUMPRODUCT('Share Calculation'!T2:T288;FiltersWeights!D2:D288)*SUMPRODUCT('Share Calculation'!T2:T288;FiltersWeights!D2:D288)/SUM(FiltersWeights!D2:D288)))/(SUM(FiltersWeights!D2:D288)-AVERAGE(FiltersWeights!D2:D288)))/SQRT(SUM(FiltersWeights!D2:D288)^2/SUMSQ(FiltersWeights!D2:D288))

Could you please explain the logic behind these calculations?
I know there are different opinions on how one should calculate standard errors in case of weighted sample, some use weighted variance and don't use effective base, some use unweighted variance and effective base. The formula doesn't resemble either of these.

I consider Sawtooth Software as one of the most pragmatic and decent teams of professionals, so I'd be happy to hear your reasoning.

Also, would it work out to simply multiply the errors by +/-1.96 to get 95% confidence interval?

Many thanks!

Igor
asked Mar 2, 2017 by IGaaa (210 points)
retagged Mar 2, 2017 by Walter Williams

1 Answer

0 votes
answered Mar 2, 2017 by Bryan Orme Platinum Sawtooth Software, Inc. (138,915 points)
...