smoothing spline based intensity estimation

The analysis is always carried out by fitting a model to part of the data (e.g., the last 30 s of a spontaneous activity recording) before "time transforming" and testing with the "extended Ogata's tests battery" the other part of the data (e.g., the first 30 s in the previous example).

The "extended Ogata's tests battery is presented in the second manuscript.

The data to which gssanova is applied are always obtained with mkGLMdf. That is, the time axis is discretized. The variable names appearing in the gssanova call are:

• event: 0 or 1 depending on the absence or presence of a spike in the corresponding bin.
• lN.x: The ellapsed time since the last spike of neuron x, when several neurons where recoded (and are processed simultaneously). Otherwise x is 1 and lN.1 is the ellapsed time since the last spike of the considered neuron.
• isiy: The duration of the yth preceding interspike interval of the considered neuron. y = 1, 2, 3, ...
• time: The time.
• some additional names corresponding to transforms of the "lN" and "isi" variables. The construction of these variables will appear explicitly in the scripts.

All times and durations are expressed in seconds.

For odor evoked activity the following "trick" is used: If the stimulus comes at sec 5.2 then a new time variable, "ot", is created where all the time values of the original time variable which are < 5 s are set to 5. In other word we force the "stimulus effect" to be homogeneous between 0 and 5. See the scripts at the bottom of the individual pages for details.