The power spectra of the MV Lyr light curves show a typical for a red noise shape: power law $P(f)\sim f^{-\gamma}$ in the high frequencies. This considerably modifies the power distribution and the use of "false alarm" probability gives an unreal high statistical significance of the detected peaks (Tremko et al. 1996). According to van der Klis (1989) the variance of the red noise power spectrum is proportional to the local mean power $\sigma_{P} \sim \langle P \rangle$ . Then a correct procedure would be to divide the PS by some mean red noise shape in order to bring the spectrum back to constant variance and then to evaluate the significance of the peaks. In the case of MV Lyr the mean red noise shape would be found through fitting the PS to function $\alpha/(1+(2\pi f\tau)^\gamma)$ ( $\alpha$ , $\tau$ and $\gamma$ are the parameters which we search for). Unfortunately, because of the large power scatter and the three unknown parameters the fitting procedure fails in the low frequencies and in practice we cannot use it.

To estimate the statistical significance of the peaks in the PS we used a rougher procedure. The PS of MV Lyr light curves are nearly constant in the frequencies approximately below 10 [c/h] (Fig. 5). This allows to apply "false alarm" probability without normalization of the power spectra in the frequency interval from 0 to 10 [c/h]. The power distribution in every PS was approximated with exponential function $\exp(-P/P_{0})$ and 99% significance levels were calculated by equation:

$\begin{displaymath} z_{0}=-P_{0}\ln(1-0.99^{1/K}).\end{displaymath}$

(1)

$K\simeq 10/\Delta f$ is the number of the independent frequencies in the interval 0-10 [c/h], where $\Delta f=1/T$ . T is the total length of the run in hours. To determine the power distribution better, the spectra were oversampled by a factor of 3. The results are shown in Fig. 2. The uncertainties included in the figure are equal to HWHM of the peaks.

$\begin{figure} \epsfig {figure=ds6230f2.eps,width=8.8cm,height=8.8cm}\end{figure}$

Figure 2: Power spectra of MV Lyr light curves (normalized to the mean power). Dashed lines mark 99% significance levels. Dash-dotted curve is the PS of a part of the U run on Jul. 19, 1993 as discussed in the text. $\ast$ marks the peaks of "50 min" QPOs below 99% significance level

In four of the runs statistical significant periodicities of about 50 min were detected. In all other runs peaks corresponding to "50 min" QPOs are seen also but they are below the 99% significance level. There are two peaks corresponding to periods about 30 min which are detected as significant also. Our opinion is that they are not due to real periodicities. The reasons for their presence can be the small length of these runs and some strong flickering peaks appearing in the minimum of "50 min" QPOs (Fig. 3). The B runs on 19 and 20 Jul., 1993 contain only two maxima of "50 min" QPOs and the flickering peaks at the minimum cause the peak in the PS corresponding to a period shorter by a factor of 2. We calculated the PS of this part of U run on Jul. 19, 1993 (with gaps introduced as in the B curve) which coincided with the B curve. From Fig. 2 it can be seen that the two PS are almost the same. Although the peaks of "50 min" QPOs in the PS of the runs on Jul. 19 and 20, 1993 are not significant, the data allow to be fitted with corresponding periods. An inspection of the results shows that the period determined from the light curves obtained simultaneously on 17 of July differ by 5.5 min, in spite of the coincidence between them. We supposed that the long gaps in the B curve are responsible for this difference mainly. The period determined from the U curve, after removing of the data corresponding to the gaps in the B curve, was 45 min. So, we decided that the value 50.0 min is more reliable and accept it as typical of QPOs during the night. It is necessary to note that the curves from July 17, 1993 show a peculiarity: every second minimum of QPOs is deeper.

To study $0{\hbox{$.\!\!^{\rm d}$}}1379$ modulations of MV Lyr brightness reported by Borisov (1992) we used composite light curves for 1993 U and B bands observations. The PS did not show significant peaks in the vicinity of the corresponding frequency. The revision of the individual light curves, including those from 1992, showed that modulations with a period of about 3 hours could be suspected in the light curve from July 18, 1993 only. From the best sinusoidal fit a period of 3.22 hours was obtained. The data from July 5, 1992 show a linear trend which may be a result of 3.3 hours modulations also.

The sinusoidal fits to the individual light curves obtained using the found periods and the amplitudes determined from these fits are shown in Fig. 3. The amplitudes of 3.3 hours modulations are 0.055 and 0.063 mag for July 5, 1992 and July 18, 1993 respectively. They are close to the mean cycle amplitude of 0.042 mag found by Borisov (1992). In his individual curves, however, the amplitudes of the signal vary from 0.018 to 0.077 mag. Skillman et al. (1995) regard their Aug./Jul. 1993 data as supportive of $0{\hbox{$.\!\!^{\rm d}$}}1379$ signal only and admitted it to be a transient one. Therefore it is possible that this signal is present with lower amplitudes in most of our curves but is obscured by large amplitude flickering and QPOs, or is not present at all.

Unfortunately, we do not have observations in at least four consecutive nights, both in U and B bands and could not confirm the four day modulations. But the three consecutive U band observations indicate an increase of the mean star's brightness with $\sim 0.06$ mag. B band observations covering descending branch of the wave, show a decrease of the brightness with $\sim 0.05$ mag.

$\begin{figure} \epsfig {figure=ds6230f3.eps,width=18cm,height=9cm}\end{figure}$

Figure 3: The light curves of MV Lyr and the best fits to them. The arrows mark the deeper minima of the QPOs discussed in the text. " $\sqcap$ " marks the strong flickering peaks appearing in some minima of the QPOs. The fits with dotted lines are performed with periods corresponding to peaks marked with $\ast$

3 Periodicity search