SPSS is probably not the best program for curve fitting relevant to your application.  I would suggest you should try some other program specifically designed for analysis of kinetics parameters in biochemistry, pharmacology, or toxicology.   For simple estimates you can use any graphic program (e.g EXL) to plot x and y response data points and fit suitable degree polynomial.  This will give you a typical dose-response curve.  EC50 is the x value for the curve point that is midway between the max and min parameters. It is called the half-maximal effective concentration. Equivalent definitions relevant to your application are ED50 (half-maximal effective dose) and for inhibition curves IC50 (half-maximal inhibitory concentration).  You can find reasonably accurate estimates of these parameters manually. 

For more complex curve fitting analysis you may consider using software program SIGMAPLOT, which is widely recognized both in science and industry for this kind of applications. You can find more specific information for your application on http://www.sigmaplot.com/products/sigmaplot/productuses/prod-uses43.php   

SPSS is probably not the best program for curve fitting relevant to your application.  I would suggest you should try some other program specifically designed for analysis of kinetics parameters in biochemistry, pharmacology, or toxicology.   For simple estimates you can use any graphic program (e.g EXL) to plot x and y response data points and fit suitable degree polynomial.  This will give you a typical dose-response curve.  EC50 is the x value for the curve point that is midway between the max and min parameters. It is called the half-maximal effective concentration. Equivalent definitions relevant to your application are ED50 (half-maximal effective dose) and for inhibition curves IC50 (half-maximal inhibitory concentration).  You can find reasonably accurate estimates of these parameters manually. 

For more complex curve fitting analysis you may consider using software program SIGMAPLOT, which is widely recognized both in science and industry for this kind of applications. You can find more specific information for your application on http://www.sigmaplot.com/products/sigmaplot/productuses/prod-uses43.php   

Hi, I can recommend two commercial software solutions (I have worked with both): Sigmaplot (see Andrews answer) and Prism (from GraphPad). Both programs offer an extensive tool collection including curve fitting. SPSS is great, but definitely not designed to adress this task. I personally prefer Prism but it may depend on individual preferences. The performance and user-friendliness is quite similar. If you prefer a non-commercial solution you will find a broad choice of reasonable alternatives by a simple google search. The calculations will be largely identical to the abovementioned software, but the graphic output will be less impressive.

You can use Dose-response curve with four parametric logistic equation using Sigma Plot. Anyway, the IC50 that you will get using this equation will be very close to the one calculated using Excel.

Disappointing answers from this group!!

How about the experimental design...are we talking about independent test organisms or are treatment combinations applied to dependent groups....Fitting the curves in these programs is fine, but don't expect to get accurate uncertainty bounds, and if there is imbalance among treatment combinations due to effects related mortality, be prepared for substantial bias in estimates.  

Is the response discrete of continuous?  Need to select appropriate general linear model, to handle counts, binomial or continuous endpoints.  Not possible in Sigma plot or Excel.   

Collapsing the data down to a handful of proportions or averages is a mistake and an old fashioned way to look at this sort of data.

Results in a dramatic loss of information and likely grossly understated uncertainty bounds.

Look for a good statistician at your university to help get this right.  The experimental design matters!!

thank you all for your answers

I'm not sure why anyone would say SPSS can't do Probit or logit analysis to assess EC50. It can do this and does so very well. See below.

Yes, SPSS Statistics can provide an estimate of the LD50 in the PROBIT procedure. The LD50 is defined as the dose required to kill 50% of exposed organisms. PROBIT produces estimates of the dose required to kill various percentages from 1-99%, with fiducial confidence intervals where they can be calculated. The values are listed under the column labeled for the dose variable in the section of the output entitled "Confidence Limits for Effective VARNAME" (VARNAME is replaced by the name of the variable).

The PROBIT procedure is in the Regression Option. To access it via the menus, specify Analyze>Regression>Probit. Note that PROBIT is designed for grouped dose-response data, and expects data to be structured in what is sometimes called R-of-N or events/trials format, with one variable holding the number of responses for a given dosage and a second variable holdings the number of trials. For a standard LD50 estimation, you would have these two variables plus a dosage variable, which would be entered as a covariate.

There is an example in the Case Studies (Help>Case Studies>Regression Option>Probit Analysis) that illustrates this, though with a marketing example, so the estimation relates to obtaining a certain percentage response to a marketing campaign rather than killing a certain proportion of organisms, but the math is the same in both cases.