Erica: Minor differences may be due to rounding error. Also, in case you're using Excel to do the actual calculations, note that many of Excel's statistical functions use notoriously bad / inaccurate algorithms (e.g. McCullough, 2005, 2008; Yalta, 2008; ). In case of the reversed sign, you have to take block order into account. I recommend you work through the various SPSS syntax procedures available for calculating D from the raw latency data.
Thanks Dave,
That is the formula I'm using. It produces D scores that are close, but not the same as the ones automatically generated. Attached is an excel spreadsheet where I replicate the computation for two examples (one producing negative Ds instead of positive ones). Am I missing something?
Erica
It's the improved scoring algorithm put forth in Greenwald et al. (2003). Also see https://www.millisecond.com/forums/Topic3444.aspx.
Originally my script did not include the automatic calculation of D, but after realizing I could do this, I used a template to add that feature.
Unfortunately now I'm stuck with about 50 subjects for whom I do not have a D. I've tried to figure out which formula was used in the newer ones that are calculated by recomputing the score using Greenwald's D formula and comparing it to the output. So far I cannot get the same Ds from my data that Inquisit is producing. Could someone tell me what formula is used in the automatic calculation of D? I just want to make sure I'm going through the same process for my initial 50 data points.
Thanks,Erica