and truncated at each end to present curves of equal length. Measurements were recorded once per day per cycle from 8days before the day of ovulation and until 15days after the ovulation.A woman may have one or several cycles. The length of the observation period is 24days Some measurements from some subjects were missing due to various reasons The data set consists of two groups :the conceptive progesterone curves (22menstrual cycles)For more detail about this
data set, see Yen and Jaffe (1991), Brumback and Rice (1998) and Fan Zhang (2000),among others
Figure 1.1 (a) presents a spaghetti plot for 22 raw conceptive progesterone curves Dots indicaed with straight line segments .The problem of missing vales is not serious here because each cyxle curve has at least 17 out of 24 measure ments. Overall, the raw curves present a similar pattern:before the ovulation day (Day 0 ), the raw curves are quit flat ,but after the ovulation day; they generally move upward.However,it is easy to see that within a cycle curve, the measurements vary qround some underlying curve which appears to be smooth, and for diffeent cycles, the underlying smooth curve are different from each other Figure1.1 (b) presents the pointwise means dot-dashed curve )with 95% pointwise standard deviation (SD)band (cross-dashed curves).They were obtained in simple way :at each distinct design time point t,the mean and standard deviation were computed using the cross-section data at t, it can be seen that the pointwise mean curve is rather smooth ,although it is not difficult to discover that there is still some noise appeared in the pointwise menn curve
Figure 1.2 (a) presents a spaghetti plot for the 69 raw nonconceptive progesterone curves.Compared to the conceptive progesterone curves.these curves behave quite similarly before the day of ovulation,but generally show a different trend after the ovulation day. It is easy to see that,like the the conceptive progesterone curves the underlying individual cycle of the nonconceptive progesterone curves appear to be smooth ,and so is their underlyling mean cuve.A naïve estimate of the underlying mean curve is the pointwise mena curve, shown as dot-dashed curve in Figure1.2(b).The 95%poinwise SD band (cross-dashed curves)provides a rough estimate
For the accuracy of the naïve estimate
The progesterone data have been used for illustrations of nonparametric regression methods by several authors ,For example, Fan ang Zhang (2000)used them to illustrate their two-stp method for estimating the underlying mean function for longitudinal data or functional data .Brumback and Rice(1998)used them to illustrate a smoothing spline mixed-effects modeling technique for estimating both mean and individual function,while WU and Zhang (2002a)used them to illustrate a local polynomial mixed-effects modeling approach.
着急需要!!!