Time-Intensity

What is Time-Intensity Analysis?

Time-Intensity (TI) is a temporal sensory method first introduced in the 30s. It became notably used in sensory analysis in the 50s (Söjström (1954)) and emerged in the 70s with the improvement of recording instrument.

During TI evaluations, assessors are asked to score the intensity of perception of a single attribute over consumption of the product. Compared to single point measurements, the analysis of the development and decline of particular sensory characteristics may reveal rich information in order to distinguish products or perceptions. This type of analysis may be applicable to a variety of product, ranging from the level of sweetness of a beverage to the feeling left by a lipstick.

TI Data usually consist in several intensity measurements scored by an assessor and recorded at several time steps. Each of those measurements should be associated to a product identifier. XLSTAT also offers the possibility to indicate an assessor as well as a session identifier.

Time-Intensity in Excel with XLSTAT

The first step of the TI analysis in XLSTAT is to measure the characteristic parameters on each temporal curve. The initial time of exposure to the stimulus is considered as the first time point on each curve and there are 10 distinct parameters defined as follows:

• I max: peak intensity or maximum observed intensity on the whole curve;
• T start: time point where the reaction to the stimulus is first perceived on the curve, defined as the first intensity value exceeding X% of the peak intensity;
• T max: time position of the peak intensity on the curve;
• T plateau: time duration around the T max where the measured intensity is greater than (100-X)% of the peak intensity;
• T ext: time point of extinction of the perception of the stimulus, defined as the position in time after the peak intensity where the measured intensity is lower than X% of the peak intensity;
• R increase: slope or rate of intensity increase between T start and T max;
• R decrease: slope or rate of intensity decrease between T max and T ext;
• Area before: the area under the curve before the peak intensity;
• Area after: the area under the curve after the peak intensity;
• Area: the total area under the curve, equal to the sum of Area before and Area after.

Where X is the value of the significance level expressed in %.

The measured curve parameters are displayed in a summary table. Time-Intensity curves are expected to match a bell shape pattern. If for some reason, the algorithm detects that one or several curves present pathological characteristics (constant intensity, several maximum, etc…), a message is displayed so that the user can investigate which curve(s) should be removed from the analysis.

The visual control of each curve is an important step in a TI analysis. The user should use its field expertise to make sure curves have meaningful characteristics. To this effect, XLSTAT offers the possibility to display all the recorded curves either on an individual chart or superimposed on a single chart to facilitate the comparison between curves.

In addition to individual time intensity curves, it is also very useful to visually summarize the panel perception of a given stimulus for different products. This can be done easily in XLSTAT by creating a synthetic curve either for the whole data set or for each product identifier. Several techniques are proposed to generate this synthetic curve:

• Average: the synthetic curve is the time step average of all individual curve;
• Parametrized: the synthetic curve is build up from the measured curve parameters;
• Overbosch method: the synthetic curve is created following the approach first proposed in Overbosch (1986);
• Liu and MacFie method: the synthetic curve is created following the approach proposed in Liu (1990).

The last two techniques require that the user specifies an additional parameter which is the desired number of bins for the synthetic curve before and after the peak intensity (the total number of bins of the synthetic curve is therefore twice that number).

Finally, an ANOVA is performed on each measured curve parameter separately to assess the product effect and possibly an assessor and or repetition effect. Depending on the selected effects, several model configurations are available to account for potential interactions between products, assessors and sessions. Furthermore, XLSTAT allows the user to treat Assessors and/or Sessions as random effect instead of fixed effect. XLSTAT provides Type III SS ANOVA tables.