Datasets:

SouthernCrossAI
/

Arxiv

Dataset card Data Studio Files Files and versions Community

Split (1)

train · 132k rows

Filename stringlengths 22 64	Paragraph stringlengths 8 5.57k
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	where n(ρ) is the measurement noise, resulting in a sin- gle covariance peak shifted by φhi. (Note the change in subscripts.) K(ρ, hi) is the theoretical spatial covari- ance of the scintillation from a single star produced by a layer at height hi where (cid:82) C 2 N (hi)dhi = 1, assuming Kolmogorov turbulence.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	such that the height of the auto-covariance peak (i.e. ∆t = 0) is approximately the height of the spatio- temporal cross-covariance peak displaced by a distance V (hi)∆t. To adequately capture motion of slowly mov- ing layers a long ∆t is required. However, long ∆t val- ues are blind to rapidly moving layers. Using multiple values of ∆t, a full V (h) proﬁle can be obtained. The algorithm used to analyse spatio-temporal covariances for V (h) proﬁles is presented in Mohr et al. (2008a).
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	The ﬁrst SCIDAR measurements taken at MJUO oc- curred in April 1999 using the system designed by Im- perial College (Johnston et al. 2002). These data re- vealed the presence of strong NGT and two diﬀerent high altitude layers located at approximately 11 and 13 km above sea level, with an estimated r0 of 12.3 cm for the full proﬁle. Temporal analysis indicated that the velocities of the 11 and 13 km layers were 6.45 and 11.63 ms−1 respectively, whereas the NGT layer was attributed to dome seeing. However the measure- ments were collected over a single observation run of 10 nights where only 50% of the nights provided useful observing conditions (Johnston 2000). For the devel- opment of AO for MJUO, it was decided that a more complete picture was required of variations in the tur- bulence proﬁle with respect to season and weather.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	SCIDAR measurements for trending purposes were collected at MJUO from 2005 2007 using UC-SCIDAR. The majority of measurements were taken using the 1-m McLellan telescope at a focal ra- tio of F/13.5. Due to the noise characteristics of the CCD cameras only a handful of double or binary star systems were suitable. The location of MJUO per- mitted the binary star systems α Cru and α Cen to be used for a large portion of the year. However in the summer months, when both α Cru and α Cen are too far from zenith, fainter star systems such as θ Eri and υ Car could be used with the latest version of UC-SCIDAR, which has more sensitive detectors. For the UC-SCIDAR system stellar separations should be limited to between 4 and 20 arcsec with the appar- ent magnitude of the primary star no fainter than 3.5, where the magnitude diﬀerence should be limited to ∼ 2.5. Table 1 gives the stellar parameters for the stars for measurements presented in this paper.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	included in the trending analysis, to ensure that turbu- lence was not subject to excessive blurring, resulting in the underestimation of layer strengths. Kl¨uckers et al. (1998) used exposure times raging from 1.6 – 2.7 ms, whereas Avila et al. (2008) report use of exposures of 3 ms. Measurements taken at zenith angles, ζ, greater than 35◦ were excluded from site proﬁling. Measure- ments with 30◦ < ζ ≤ 35◦ were included only if sup- porting data taken at ζ ≤ 30◦ was acquired within approximately 30 minutes of the run. This was to en- sure that a suitable altitude range was sampled and that measured NGT was not subject to vertical air ﬂows located near ground. A signiﬁcant portion of the data collected during 2006 was subject to corrupt CCD readout and hence was not included in this study.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Each observation period usually consisted of three to four consecutive nights. Data sequences typically consisted of 5000 frames from each camera which was recorded into ﬁle blocks of 500 frames per ﬁle. To decrease the processing time, 2D spatio-temporal co- variances for each camera were calculated using 2500 frames with time delays between consecutive frames of ∆t = 0, dt, 2dt . . . 6dt, where dt represents the frame rate of the CCD camera used. This ensured that at least 1000 cross-correlations were used in the longer ∆t ensembles. It should be noted that all available frames were used such that for a ∆t = 3dt, frame 1 was correlated with frame 4, frame 2 with frame 5, frame 3 with frame 6, and so on.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Measured velocities were cross-checked between se- quential runs and the various ∆t covariances, and layer heights from temporal analysis were cross-checked with the corresponding C 2 N (h) analysis for a given run to eliminate any falsely detected layers. Data collected in 2007 was collected at 60 Hz, whereas data collected prior to 2007 was collected at 30 Hz due to a limitation in the CCD cameras used at the time.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	In a signiﬁcant number of cases there was simulta- neous pupil-plane and generalised SCIDAR data. This provided an added check to layer heights in the free at- mosphere, as well as an insight into how the strong low altitude layers can aﬀect the measurements obtained for high altitude layers.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	NGT can blur any covariance peaks detected during temporal analysis, particularly when using star sys- tems with a narrow angular separation. As such it is not always possible to determine the height of layers using temporal analysis as the primary and secondary peaks blur together.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	N (h) proﬁles presented in this paper include dome/mirror seeing as any AO system developed for the McLellan 1-m telescope at MJUO would need to compensate for dome/mirror seeing too.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	calculated from UC-SCIDAR data have been determined for a wave- length, λ, of 589 nm, although the measurements were collected using broadband white light. The variable weather at MJUO (ranging from calm, clear nights to gusting winds and thickening clouds) resulted in a variety of proﬁles being detected. During increas- ing cloud cover, with moderate to high ground wind speeds, the NGT present was exceptionally strong re- sulting in a pupil-plane r0 that was similar to that found for the corresponding generalised data during analysis. In these cases the generalised r0 estimate was used in the determination of the pupil-plane av- erages to remove any bias toward noise in pupil-plane results. Results are presented from the most recent to the earliest.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figures 4 and 5 show the pupil-plane and generalised C 2 N (h) proﬁles respectively for data collected in 2007. The NGT measurements at MJUO tend to dominate and typically mask any activity present in the upper layers, as seen in Figures 4(a) and 5(a). To reveal possible features in these upper layers the C 2 N (h) pro- ﬁles are scaled so that the colour range is limited to N (h)∆h values between 10−14 and 10−13 m1/3. Val- C 2 ues below 10−14 m1/3 are likely to result in images that are diﬀraction-limited rather than turbulence-limited on a 1-m telescope. r0 is 1.55 m for a layer with N (h)∆h = 10−14 m1/3 and a wavelength of 589 nm. C 2 Values greater than 10−13 m1/3 are approaching lev- els that are classiﬁed as strong turbulence (Andrews 2004). Figure 4(b) shows the colour scaled images for pupil-plane measurements from 2007. Figure 5(b) shows the colour scaled proﬁles for 2007 generalised measurements.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figure 4: Pupil-plane C 2 N (h) proﬁle trends ob- served over 2007. Gaps have been added where no data is present for more than two hours. Data from the individual months is separated by solid black lines. The image in (b) has been scaled such that N (h)∆h value above 10−13m1/3 is set to the any C 2 maximum colour range and any C 2 N (h)∆h value below 10−14m1/3 is set to the minimum colour range.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	level does not match the strength of the layer detected at comparable height in pupil-plane data.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	From equations (4) and (6), there is an assump- tion that each layer of turbulence present above a site is statistically independent and hence the strength of the layers found is not dependent on other layers in the structure. If this assumption holds true then the measured strength of the detected turbulence from any given high altitude layer should be the same regard- less of whether pupil-plane or generalised measure- ments were employed. However issues arise under con- ditions of medium to strong NGT resulting in an un- derestimate of the strength of turbulence located in the higher levels in generalised SCIDAR measurements (Mohr 2009). A correction factor can be applied to the high altitude layers detected in generalised data.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figure 5: Generalised C 2 N (h) proﬁle trends ob- served over 2007. Time and C 2 N (h) colour scal- ing used is as per Figure 4. (c) Correction factors applied to turbulent layers detected in the free at- mosphere is incorporated. See text for details.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figure 6: Pupil-plane and generalised C 2 N (h) pro- ﬁle trends observed over 2005. Time and C 2 N (h) scaling, and generalised data corrections used are as per Figure 5.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	ulations (Mohr 2009). It should be noted that this correction has little eﬀect on the estimated r0 for the proﬁles, due to the strength of the NGT layer that dominates the proﬁles. Following the correction it can be noted that similar strength turbulent layers are seen in both the pupil-plane and generalised SCIDAR data for the high altitudes (Figures 4(b) and 5(c)). Where possible a similar correction has been applied to all generalised data presented in this paper.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	The layer found at 12 – 14 km above sea level can be associated with turbulence found in the tropopause region, which is commonly incorporated into models for C 2 N (h) proﬁles (Hardy 1998). For a site such as MJUO, a signiﬁcant level of turbulence at low alti- tudes is expected. MJUO is located roughly 50 km west of the Southern Alps. With prevailing westerlies over much of New Zealand, the low- to mid- altitude wind structure is signiﬁcantly aﬀected by the terrain associated with the Southern Alps (Sturman and Tap- per 1996).
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	For data collected in 2005, shown in Figure 6, a strong low altitude layer located at less than 5 km above sea level is visible. Also present is a weaker high altitude layer that ranges between 10 – 14 km above sea level. In some months an additional layer can be seen at 6 – 8 km above sea level.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figure 7: Pupil-plane and generalised C 2 N (h) pro- ﬁle trends observed during autumn from 2005 to 2007. Time and C 2 N (h) scaling, and generalised data corrections used are as per Figure 5.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	N (h) proﬁles obtained for the winter months (i.e. June and July) with UC- SCIDAR collected in both 2005 and 2007. A high al- titude layer is seen consistently at approximately 12 km above sea level with an average C 2 N (h)∆h strength of approximately 3 × 10−14 m1/3. This is of similar strength to that seen in the autumn months (i.e. April and May), suggesting little to no diﬀerence in the high altitude layer. However, an additional mid-altitude layer is seen in some data, with heights ranging from 6 – 8 km above sea level and with varying strength. As seen with the April/May data, low altitude turbulence extends up to 5 km above sea level. Again little can be ascertained from the generalised data about trends in the heights and strengths of the high altitude layers.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figure 8: Pupil-plane and generalised C 2 N (h) pro- ﬁle trends observed during winter from 2005 to 2007. Time and C 2 N (h) scaling, and generalised data corrections used are as per Figure 5.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	providing a measure of the eﬀective height of dom- inant turbulence if the multi-layer structure was re- placed with a single layer. Note that h0 values shown reﬂect the distances above sea level. For generalised data, h0 has been corrected for defocus distances for each observation sequence.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	January 2007 data exhibited similar r0 and θ0 val- ues for both pupil-plane and generalised data, which indicates a signiﬁcant low-altitude layer. This is also shown in the h0 values of 4.8 and 2.9 km for pupil- plane and generalised data respectively. In summer, the longer, warmer days heat the surrounding ground and buildings which can lead to greater NGT eﬀects. The pupil-plane θ0 values obtained for May are smaller than those obtained for June although the r0 values are similar for the two months. The diﬀerence in θ0 can be attributed to the high altitude layers found in May, which were stronger and higher than those found in June (Figure 5(b)). This is also seen in the higher pupil-plane h0 value for 2007 May of 7.3 km compared to 5.7 km for 2007 June.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	trend to that of r0, with the winter months having a value of approximately 2 arcsec compared to 1.3 – 1.7 arcsec for autumn and spring. θ0 for 2005 March was 1.6 arcsec with a large standard deviation σθ0 of 0.7. This is associated with the noise present in a sig- niﬁcant portion of the runs during 2005 March. The r0 values for generalised measurements are reasonably consistent throughout the two years; 5 – 6 cm for 2007, and 7 cm for 2005. Little variation is seen for gener- alised θ0 values across the entire campaign, suggesting a consistent, dominating NGT layer. The signiﬁcant variations seen in pupil-plane r0 and θ0 values suggest that the strengths of the high altitude layers ﬂuctuate. The average r0 for all UC-SCIDAR data was 12 ± 5 cm and 7 ± 1 cm for the pupil-plane and generalised measurements respectively. θ0 was 1.5 ± 0.5 arcsec and 1.0±0.1 arcsec for the pupil-plane and generalised mea- surements respectively. h0 was 6 ± 1 km and 2.0 ± 0.7 km for pupil-plane and generalised measurements re- spectively. The values of h0 obtained does reinforce that the dominating turbulence at MJUO is located at near ground altitudes.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	UC-SCIDAR estimates for r0 are smaller than those obtained in 1999 April using the Imperial College sys- tem (i.e. 12.3 ± 1.4 cm) (Johnston et al. 2002). Al- though layer height estimates were similar for the two systems, the strength of the NGT layer was in the or- der of 6 – 10 times stronger for UC-SCIDAR data. Some variation can be expected due to the amount of time that has passed between the two instruments be- ing used. However the lower r0 values obtained using UC-SCIDAR do match the observation conditions typ- ically seen by observers, which has a nominal angular resolution, θres, of ∼ 2 arcsec (A. Gilmore (MJUO) 2006, private communication). θres from UC-SCIDAR was 2.5 arcsec for the full proﬁle, calculated at a wave- length of 589 nm. θres for data taken in 1999 April was 1.2 arcsec.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	The large variation in the pupil-plane measure- ments suggests not only a possible relationship with seasonal changes, but also with the weather at the site. June and July of 2005 saw high pupil-plane r0 and θ0 values which can be attributed to the calmer weather seen during these observational periods. However most other months show similar r0 and θ0 estimates within the margin of error. This suggests that weather con- ditions have a greater inﬂuence on the proﬁles ob- tained. It is suggested that data obtained from UC- SCIDAR be correlated to meteorological data to inves- tigate whether SCIDAR data could predict weather- related seeing over the site.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Table 2: Monthly averages and standard deviations for r0, θ0 and h0 for all months. h0 values are indicated as distances above the sea level. Elevation of MJUO is 1024 m. Generalised values have been computed based on corrected proﬁles. Generalised values of h0 have been corrected for defocus distances.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Table 3: Nightly averages for fG for all months. Generalised values have been computed based on corrected proﬁles.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figure 9: Average wind speeds, \|V (h)\|, for obser- vations taken during 2007. Measurements from the individual months are separated by solid black lines.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	niﬁcant NGT, a low-altitude layer was seen with wind speeds ranging between 10 to 24 ms−1.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Velocity measurements obtained in April 1999 in- dicated high-altitude layers at 11 and 13 km travelling at 6.45 and 11.63 ms−1 respectively. This is consistent with the measurements obtained using UC-SCIDAR for layers at this altitude.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Table 3 shows the monthly averages and standard deviations for fG. Note that the value shown for 2007 January is averaged across both the pupil-plane and generalised data. The values for 2007 June reﬂect the average for data collected on May 31 and June 1 only. The monthly calculated Greenwood frequency av- erages, fG, for 2007 were approximately 30 Hz or less. For a wavelength of 589 nm, the estimate for fG ranges between 30 – 90 Hz depending on the models used for C 2 N (h) and V (h) proﬁles (Tyson & Frazier 2004). The measured fG obtained at MJUO is at the bottom end of this range. It is most likely that the fG has been underestimated due to the gaps that exist in the V (h) proﬁles.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	variance strengths in the measurement plane being suf- ﬁciently strong with respect to the background covari- ance noise. Aperture normalisation in the data am- pliﬁes noise, especially near the aperture edge. Peaks approaching the aperture edge can be hidden by the noise. The detection of only the central peak with one secondary peak, termed partial triplet analysis (Mohr et al. 2008a), does allow for more layer velocities to be found. In addition, some covariance peaks may be obscured due to their close proximity to other peaks. Although all layers seen in V (h) will have an associ- ated C 2 N (h) layers will have a measurable V (h) due to the position of the covariance peaks relative to other peaks and the aperture edge and the resulting covariance strength. This results in gaps in the measured V (h) proﬁle and is reﬂected in the large σfG values.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figure 11: As per Figure 9, except for observations taken during winter months of 2005 and 2007.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	low- to mid-altitude layers, however high-altitude layer velocities may not be measurable. Temporal analysis was performed on data from June and July 2005 data only. These months provided the most reliable data set with a large number of runs utilising exposures of 1 – 2 ms.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figure 10 shows the average wind speeds for 2005 June and July. In June a layer at approximately 12 km above sea level was moving consistently at 12 – 15 ms−1. In July there was much more scatter in the average velocity.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Using the V (h) proﬁles obtained for 2005 June and July, the average Greenwood frequency, fG, was found to range between 10 and 20 Hz. There are signiﬁcant gaps present in the V (h) proﬁles, as indicated by the standard deviation in fG, σfG , ranging between 5 and 15 Hz (Table 3). As such fG is likely to be underesti- mated.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figure 11 shows the average wind speed for the winter months (June and July). Although a signiﬁcant amount of scatter exists for the data from 2005 July and 2007 June the layer heights found are similar. The high altitude layer has an average speed of 10 – 15 ms−1 in calmer weather, but speeds in excess of 25 ms−1 at other times.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	around a mean value (Hardy 1998). Commonly used models represent a mean proﬁle of C 2 N (h) measure- ments taken over extended periods of time.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	The Hufnagel-Valley (HV) model is commonly used to describe the average turbulence at an astronomical site (Tyson & Frazier 2004). A standard HV model consists of three main components: an exponentially decreasing C 2 N (h) through the troposphere; a peak at approximately 10 km above ground corresponding to a tropo-pause layer; and a strong surface layer (Hardy 1998).
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	where A is the turbulence coeﬃcient for near-ground turbulence (i.e. ∝ C 2 N (0)) and HA is the height for its 1/e decay, B and HB are similarly deﬁned for turbu- lence in the troposphere, and C and HC are related to the turbulence peak located at the tropopause. The fourth term in equation (10) can be used to deﬁne one or more isolated layers, where D and HD deﬁne the strength and height of the layer and d speciﬁes the layer thickness.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	was found at heights ranging from 6 – 8 km above sea level. MJUO3 incorporates an additional layer at 6.5 km above sea level, and is the recommended model for use in the AO design for MJUO. The proﬁles of the HV 5-7 model, MJUO1 and MJUO3 are shown in Figure 13.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Additional layers at approximately 2 and 6.5 km about sea level have also be regularly detected at other sites. Avila et al. (2001) reported the detection of ﬁve diﬀerent layers during the 1998 site testing campaign of Cerro Tololo and Cerro Pach´on in Chile. Strong layers were found at the tropopause region (i.e. 11 – 12 km) and at low altitudes, extending up to approximately 4 km above sea level. Although the layer found at 6.5 km was often weaker than the low altitude turbulence, it was consistently present. Similar low- to mid- altitude turbulence was detected at San Pedro M´artir, Mexico (Avila et al. 2007).
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	The HV model is such that the resulting peak present at the tropopause region is not a spike but rather a broad peak, which accounts for variations in layer heights and ﬂuctuations in the turbulence strength seen over time. Instantaneous spikes and variations in the heights of the turbulent layers are smoothed out. This has the eﬀect of broadening the peaks seen, par- ticularly in the tropopause region. Although it would be ideal to reﬁne the model such that the breath of the tropopause region is not so wide, as a model for AO design MJUO3 is the recommended model. Using a wavelength of 589 nm, r0 is estimated to be 6 cm for MJUO3. θ0 is estimated to be 0.9 arcseconds. h0 is estimated at 4.2 km above the measurement plane. If a defocus distance of 3 km was used the resulting h0 of 1.2 km would be in line with the values obtained from the generalised UC-SCIDAR measurements.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figure 12: C 2 N (h) model ﬁtting. Both (a) and (b) show selected pupil-plane and generalised data from ((cid:70)) 2007 June, ((cid:13)) 2007 May, ((cid:3)) 2005 June and (♦) 2005 April. Models shown are (—) the HV 5-7 model, (– –) MJUO1: a modiﬁed HV model and (– · –) MJUO3: a modiﬁed HV model that incorporates two additional layers.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figure 13: C 2 N (h) turbulence models for MJUO. (—) HV 5-7 model; (– –) MJUO1: Modiﬁed HV model; (– · –) MJUO3: Modiﬁed HV model in- corporating two additional layers. MJUO3 is the recommended model.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	aFor all models indicated A = 17 × 10−15, HA = 100 m, B = 27 × 10−17, and HB = 1500 m. br0 and θ0 values are speciﬁed for λ = 589 nm for the full proﬁle and for h > 3 km above the telescope in brackets. cMJUO2 includes a strong low-altitude layer. dMJUO3 incorporates an additional mid-altitude layer for which the parameters are indicated in brackets.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figure 14: Wind speed analysis for observations made on 2007 May 3. Models shown are (- -) the standard Bufton model, (—) a modiﬁed Bufton model with HT = 11 km, and (- -) a modiﬁed Bufton model with HT = 11 km and ζ = 20◦. Elevation for MJUO is 1024 m above sea level.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	layers is dependent on the velocity detected near the ground. Figure 15(a) shows the instantaneous wind speeds obtained during June and July of 2005 and May and June of 2007. Overlaid are four diﬀerent V (h) models developed to encompass the range of velocity characteristics detected. The parameters for the four models for MJUO are listed in Table 5.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	where V (H1) is the velocity of a low-altitude layer lo- cated at H1 above the telescope with a thickness of L1.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Figure 15: Fit of V (h) models to measured pro- ﬁles. Models shown are (—) MJUO1V, (– –) MJUO2V, (· · · ) MJUO3V and (– · –) MJUO4V.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	tered and hence should be the preferred model for AO design. MJUO4V was developed for situations where high ground wind speeds are present.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	N (h) model and the MJUO3V V (h) model for moderate ground wind speeds, fG is es- timated to be 79 Hz for a wavelength of 589 nm, which is more in line with the 30 – 90 Hz range suggested by Tyson & Frazier (2004).
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Although adequate models for V (h) proﬁles were determined, it is recommended that more temporal data be collected to reﬁne the models. An investiga- tion into the correlation between low and high altitude velocities could also be conducted.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	UC-SCIDAR measurements taken between 2005 – 2007 detected strong NGT with a weaker layer located at 12 – 14 km above sea level. On calm nights a third mid- altitude layer was detected at ∼ 6 km above sea level. In a signiﬁcant amount of data strong low altitude tur- bulence extended up to 5 km above sea level. The measurements suggest an average r0 of 12 ± 5 cm and 7±1 cm for pupil-plane and generalised proﬁles respec- tively. This corresponds to an angular resolution, θres, of 2.1 arcsec for the full proﬁle. The average θ0 values were 1.5±0.5 arcsec and 1.0±0.1 arcsec for pupil-plane and generalised proﬁles respectively. Average h0 val- ues were 6 ± 1 km and 2.0 ± 0.7 km for pupil-plane and generalised measurements respectively.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	plied to generalised proﬁles to the layers detected in the free atmosphere in the post-processing phase. It would be preferable to have these corrections incorpo- rated into the inversion algorithm to reduce the amount of double handling of data. To create an automatic de- tection system for this, a thorough investigation as to the eﬀects of NGT strength on the detected strength from higher layers should be conducted, including var- ious altitudes and high altitude turbulence strength.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	Temporal analysis detected layers located at sim- ilar altitudes, with tropopause layer velocities of 12 – 30 ms−1, dependent on weather conditions. Low al- titude turbulence layers had velocities ranging from 2 ms−1 to well over 24 ms−1. No trends could be estab- lished for the values of fG, due to the gaps in the V (h) proﬁles. Little seasonal variation was detected in the C 2 N (h) and V (h) proﬁles were highly dependent on the weather conditions.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	A modiﬁed Hufnagel-Valley (HV) model was de- veloped to describe the C 2 N (h) proﬁles, incorporating a strong NGT layer, a layer at 11 km above the tele- scope (i.e. 12 km above sea level) and two additional layers: one at 5.5 km above the telescope and the other at 1.5 km above the telescope extending up to 4 km. The resulting model estimates an r0 of 6 cm for a wave- length of 589 nm, which corresponds to a θres of 2.5 arcsec. θ0 is estimated at 0.9 arcsec.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	A series of V (h) models were developed, based on the Greenwood wind model with an additional Gaus- sian peak located at low altitudes to model the V (h) proﬁles seen at MJUO. The models correspond to calm, light, moderate and strong ground wind speed condi- tions seen at the site. Using the modiﬁed HV model for C 2 N (h) proﬁles and the suggested model for V (h) pro- ﬁles in the presence of moderate ground wind speeds, fG was estimated at 79 Hz for a wavelength of 589 nm.
Processed_Optical_Turbulence_Measurements_and_Models_for_Mou.txt	The authors would like to thank C. Clare Worley (for- mer student at University of Canterbury, now at Ob- servatoire de la Cote d’Azur) for collecting the data during 2005. A signiﬁcant portion of this research used this data. Thank you to Steve Weddell (University of Canterbury) for the loan of the cameras used in the current UC-SCIDAR system. J.L. Mohr also acknowl- edges Dr Charles Jenkins and Dr Andrew Lambert for comments on her PhD thesis in which this data was presented.
Processed_The_scaling_infrared_DSE_solution_as_a_critical_en.txt	Both regular (the zero-momentum ghost dressing function not diverging), also named decoupling, and critical (diverging), also named scaling, Yang-Mills propagators solutions can be obtained by analyzing the low-momentum behaviour of the ghost propagator Dyson-Schwinger equation (DSE) in Landau gauge. The asymptotic expression obtained for the regular or decoupling ghost dressing function up to the order O(q2) ﬁts pretty well the low-momentum ghost propagator obtained through the numerical integration of the coupled gluon and ghost DSE in the PT-BFM scheme. Furthermore, when the size of the coupling renormalized at some scale approaches some critical value, the PT-BFM results seems to tend to the the scaling solution as a limiting case. This critical value of the coupling is compared with the lattice estimate for the Yang-Mills QCD coupling and the latter is shown to lie much above the former.
Processed_The_scaling_infrared_DSE_solution_as_a_critical_en.txt	The low-momentum behaviour of the Yang-Mills propa- gators derived either from the tower of Dyson-Schwinger equations (DSE) or from Lattice simulations in Landau gauge has been a very interesting and hot topic for the last few years. It seems by now well established that, if we assume in the vanishing momentum limit a ghost dressing function behaving as F(q2) ∼ (q2)a F and a gluon propagator as D (q2) ∼ (q2)a G−1 (or, by following a notation commonly used, a gluon dressing function as G(q2) = q2D (q2) ∼ (q2)a G), two classes of solutions may emerge (see, for instance, the discussion of refs. [1, 2]) from the DSE: (i) those, dubbed “decoupling”, where a F = 0 and the suppression of the ghost contribution to the gluon propagator DSE results in a massive gluon propagator (see [3, 4] and references therein); and (ii) those, dubbed “scaling”, where a F 6= 0 and the low- momentum behaviour of both gluon and ghost propaga- tors are related by the coupled system of DSE through the condition 2a F + a G = 0 implying that F 2(q2)G(q2) goes to a non-vanishing constant when q2 → 0 (see [5, 6] and references therein).
Processed_The_scaling_infrared_DSE_solution_as_a_critical_en.txt	approaches like the infrared mapping of lf 4 and Yang- Mills theories in ref. [14] or the massive extension of the Fadeev-Popov action in ref. [15] appear to point to.
Processed_The_scaling_infrared_DSE_solution_as_a_critical_en.txt	as can be obtained from Eqs.(3,5). The case a F = 0 cor- responds to the so-called decoupling solution, where the zero-momentum ghost dressing function reaches a non- zero ﬁnite value and eq. (5) provides us with the ﬁrst asymptotic corrections to this leading constant. This sub- leading correction is controlled by the zero-momentum value of the coupling deﬁned in eq. (6), which is an ex- tension of the non-perturbative effective charge deﬁni- tion from the gluon propagator [20] to the Taylor ghost- gluon coupling [21].
Processed_The_scaling_infrared_DSE_solution_as_a_critical_en.txt	TABLE 1. Gluon masses and the zero- momentum non-perturbative effective charges, taken from ref. [17] and ob- tained as discussed in the text.
Processed_The_scaling_infrared_DSE_solution_as_a_critical_en.txt	In ref. [17], the solutions of the coupled DSE system in the PT-BFM scheme (with H1 = 1 for the ghost-gluon vertex), numerically integrated for many values of the coupling at the renormalization point m 2 as a boundary condition, were studied in the light of the analytical results above presented. Here we will shortly discuss the results of this work.
Processed_The_scaling_infrared_DSE_solution_as_a_critical_en.txt	The numerical results of the PT-BFM coupled DSEs were shown in ref. [17] to behave asymptotically as eq. (4) predicts for the decoupling DSE solutions. In- deed, as the gluon propagator solutions in the PT- BFM scheme result to behave as massive ones, the eqs. (3,5) must account for the low-momentum be- haviour of both gluon propagator and ghost dressing function with H1 = 1 and a T (0) given by eq. (6), with R(m 2)/(4p ) being ﬁxed, as a boundary con- a T (m 2) = g2 dition for the numerical integration of the coupled DSE for each particular solution of the family (see tab. 1). Fur- thermore, the zero-momentum values of the ghost dress- ing function, FR(0) and of the gluon propagator, D R(0), can be taken from the numerical solutions of the DSE (for any value of the a (m = 10GeV)). These altoghether with the gluon masses obtained by the ﬁt of eq. (3) to the numerical DSE gluon propatator solutions (see the left plot in ﬁg. 1, for a (m ) = 0.16, and the results for a (m ) = 0.15, 0.16, 0.17 in tab. 1, taken from ref. [17]), provide us with all the ingredients to evaluate, with no unknown parameter, eq. (5).
Processed_The_scaling_infrared_DSE_solution_as_a_critical_en.txt	Indeed, the expression given by eq. (5) can be succes- fully applied to describe the solutions all over the range of coupling values, a (m ), at m = 10 GeV (provided that they are not very close of the critical coupling that will be deﬁned in the next subsection). This can be seen, for instance, for a = 0.16, 0.18, in the right plots of ﬁg. 1 and it was also shown for a = 0.15 in ref. [17].
Processed_The_scaling_infrared_DSE_solution_as_a_critical_en.txt	FIGURE 1. Gluon propagator (left) and ghost dressing function (right) after the numerical integration of the coupled DSE system for a (m = 10GeV) = 0.16, 0.17 taken from [17] . The curves for the best ﬁts to gluon propagator and ghost propagator data explained in the text appear as red dotted lines. the same for the black dotted line in the lefthand plot but retaining only the logarithmic leading term for the asymptotic ghost dressing function by dropping the −11/6 away.
Processed_The_scaling_infrared_DSE_solution_as_a_critical_en.txt	and where we let a crit be a free parameter to be ﬁt- ted by requiring the best linear correlation for log[F(0)] in terms of log[a crit − a ]. In doing so, the best corre- lation coefﬁcient is 0.9997 for a crit = 0.1822, which is pretty close to the critical value of the coupling above which the coupled DSE system does not converge any more, and k (m 2) = 0.0854(6). This can be seen in ﬁg. 2, where the log-log plot of FR(0) in terms of a crit − a is shown and the linear behaviour with negative slope corresponding to the best correlation coefﬁcient strik- ingly indicates a zero-momentum ghost propagator di- verging as a → a crit. Nevertheless, no critical or scal- ing solution appears for the coupled DSE system in the PT-BFM, although the decoupling solutions obtained for any a < a crit = 0.1822 seem to approach the behaviour of a scaling one when a → a crit. This is again well un- derstood in ref. [17], where the gluon propagators ob- tained from the coupled DSE system in PT-BFM were also found to obey the same critical behaviour pattern as the ghost propagator, when approaching the critical value of the coupling. Indeed, when approaching the crit- ical value of the coupling, the gluon propagators obtained from the coupled DSE system in PT-BFM must be also thought to obey the same critical behaviour pattern as the ghost propagator. In the PT-BFM, the value at zero- momentum being ﬁxed by construction [4], one should expect that, instead of decreasing, the gluon propagator obtained for couplings near to the critical value increases for low momenta: the more one approaches the critical coupling the more it has to increase. This is indeed the case, as can be seen in ﬁg. 2(b). This also implies that, near the critical value, the low momentum propagator does not obey eq. (3) and that consequently eq. (5) does not work any longer to describe the low momentum ghost propagator.
Processed_Memetic_Artificial_Bee_Colony_Algorithm_for_Large-.txt	tive function. Note that large-scale refers to a huge number of variables, i.e. n ≥ 1, 000.
Processed_Memetic_Artificial_Bee_Colony_Algorithm_for_Large-.txt	where fi, fworst and fbest denotes the nectar amounts of i-th, worst and best food sources.
Processed_Memetic_Artificial_Bee_Colony_Algorithm_for_Large-.txt	tions. The goal of this exploration is to bring a candidate solution into the local optimum.
Processed_Memetic_Artificial_Bee_Colony_Algorithm_for_Large-.txt	In the case that the solution is the global optimum, the search needs to be ﬁnished.
Processed_Memetic_Artificial_Bee_Colony_Algorithm_for_Large-.txt	On the other hand, the exploration is blind without any history information explored so far.
Processed_Memetic_Artificial_Bee_Colony_Algorithm_for_Large-.txt	Note that the separability of a function determines how diﬃcult the function is to solve.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	Univariate regression models have rich literature for counting data. However, this is not the case for multivariate count data. Therefore, we present the Multivariate Gener- alized Linear Mixed Models framework that deals with a multivariate set of responses, measuring the correlation between them through random eﬀects that follows a multivari- ate normal distribution. This model is based on a GLMM with a random intercept and the estimation process remains the same as a standard GLMM with random eﬀects integrated out via Laplace approximation. We eﬃciently implemented this model through the TMB package available in R. We used Poisson, negative binomial (NB), and COM-Poisson distri- butions. To assess the estimator properties, we conducted a simulation study considering four diﬀerent sample sizes and three diﬀerent correlation values for each distribution. We achieved unbiased and consistent estimators for Poisson and NB distributions; for COM- Poisson estimators were consistent, but biased, especially for dispersion, variance, and correlation parameter estimators. These models were applied to two datasets. The ﬁrst concerns a sample from 30 diﬀerent sites collected in Australia where the number of times each one of the 41 diﬀerent ant species was registered; which results in an impressive 820 variance-covariance and 41 dispersion parameters being estimated simultaneously, let alone the regression parameters. The second is from the Australia Health Survey with 5 response variables and 5190 respondents. These datasets can be considered overdis- persed by the generalized dispersion index. The COM-Poisson model overcame the other two competitors considering three goodness-of-ﬁt indexes, AIC, BIC, and maximized log- likelihood values. As a result, it estimated parameters with smaller standard errors and a greater number of signiﬁcant correlation coeﬃcients. Therefore, the proposed model is capable of dealing with multivariate count data, either under- equi- or overdispersed responses, and measuring any kind of correlation between them taking into account the eﬀects of the covariates.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	It is well known that the Poisson distribution is the most popular model to deal with count data under the framework of the generalized linear models (GLM) (Nelder and Wedderburn 1972). However, it is limited to equidispersed count data, i.e., when the mean of the response variable is equal to the variance. As alternatives to the Poisson model, the statistical literature for univariate count data is rich, either from overdispersed or underdispersed perspectives. For example, negative binomial (NB) type II distribution parameterized on mean and dispersion parameter known as the quadratic parametrization of the variance (Winkelmann 2008), hurdle and zero-inﬂated models (Zeileis, Kleiber, and Jackman 2008) and mixed Poisson regression (Winter and B¨urkner 2021) are frequent options to model overdispersed count data; Gamma Count (Zeviani, Ribeiro Jr, Bonat, Shimakura, and Muniz 2014), Conway-Maxwell-Poisson (COM-Poisson) (Shmueli, Minka, Kadane, Borle, and Boatwright 2005) and the Extended Poisson Tweedie (Bonat, Jørgensen, Kokonendji, Hinde, and Dem´etrio 2018) based on the Poisson Tweedie distribution (El-Shaarawi, Zhu, and Joe 2011; Jørgensen and Kokonendji 2016) can handle all three situations (under-, equi-, over-dispersed), even though the compu- tation of their probability mass function (pmf) relies on numerical methods.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	On the other hand, the bibliography for multivariate data is scarce. However, there is an increasing demand from researchers to analyze datasets with over one response variable. The beneﬁt of it is to better investigate the relationship between the response variables (Bonat and Jørgensen 2016); at the same time that numerical methods can take advantage of it once more data is available to estimate model parameters.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	Some methodologies to deal with multivariate data have been proposed and introduced by Winkelmann (2008) with a focus on building multivariate distributions. However, it comes with the price of some practical limitations. The multivariate Poisson-lognormal regression (MPLR) and the latent Poisson-normal regression models admit only overdispersed data. The multivariate Poisson-gamma mixture model (MPGM) and multivariate NB model (MNBM) are suitable only for overdispersed data with positive correlations. Another example is the copulas framework, which allows the building of multivariate distributions. But a negative correlation between many response variables is diﬃcult to model, especially for count data Inouye, Yang, Allen, and Ravikumar (2017) proposed (Nikoloulopoulos and Karlis 2009). three alternatives to deal only with equi and overdispersed data constructed either via copulas or a mixture of distributions.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	Distributions with no practical limitations were also developed. Famoye (2015) proposed a multivariate generalized Poisson regression model that can deal with any kind of disper- sion and correlation with an estimation based on the maximum likelihood (ML) paradigm. In a similar way, Mu˜noz-Pichardo, Pino-Mej´ıas, Garc´ıa-Heras, Ruiz-Mu˜noz, and Gonz´alez- Regalado (2021) proposed a multivariate conditional Poisson regression model, where the relationship between response variables is measured by a coeﬃcient in the linear predictor. In its turn, the dependence between response variables is conditional on the other response variables.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	A very ﬂexible modeling framework based on estimating functions, the Multivariate Covari- ance Generalized Linear Models (MCGLM) can also ﬁt such data (Bonat 2016). It uses only second-order moments assumptions and they estimated the parameters based on quasi- likelihood (Wedderburn 1974). It can accommodate correlated data based on an approach similar to the generalized estimating equations (GEE) (Liang and Zeger 1986), allowing both multivariate responses and correlated data. We can also cite methodologies based on Bayesian inference, such as MCMC Generalized Linear Mixed Models via MCMCglmm package (Hadﬁeld 2010) and Bayesian Regression Models using Stan - brms package (B ¨Urkner 2018), which comes with the price of a greater computational time. We are not going to discuss Bayesian models in this article.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	of correlated variables or individuals via a random eﬀect, an unobserved variable, that can follow any distribution. When the distribution of the random eﬀect is Gaussian, we have the Generalized Linear Mixed Models (GLMM). However, GLMM is widely known and used to model the correlation between sample units, not for response variables, such a method is implemented in consolidated packages in software R (R Core Team 2020), such as glmmTMB (Brooks, Kristensen, van Benthem, Magnusson, Berg, Nielsen, Skaug, Maechler, and Bolker 2017), lme4 (Bates, M¨achler, Bolker, and Walker 2015) and nlme (Pinheiro, Bates, DebRoy, Sarkar, and R Core Team 2017).
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	In this article, we propose to model multivariate count data under the framework of GLMMs to accommodate the correlation between response variables. Parameter estimates are obtained through the maximum likelihood method (Aldrich et al. 1997). We use multivariate normally distributed random eﬀects to accommodate the correlation between response variables. The estimation process is similar to the one for GLMM and was implemented in R (R Core Team 2020) through TMB package (Kristensen, Nielsen, Berg, Skaug, and Bell 2016). Here, we will focus on multivariate overdispersed data, whereas da Silva, Laureano, Petterle, J´unior, and Bonat (2022) presented a similar approach focused on underdispersed data. This modeling approach for multivariate count data is evaluated for three diﬀerent distributions of the response variables: Poisson, NB, and COM-Poisson mean parameterized (Huang 2017). Even though COM-Poisson does not belong to the exponential family, we refer to it as a GLMM framework, once the estimation process remains the same regardless of the distribution being used.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	This article contains six sections, including this introduction. Section 2 describes the datasets used to provide illustrative applications of the model. Section 3 proposes the multivariate generalized linear mixed model (MGLMM) model. Section 4 shows the result of the simulation study to assess the estimators’ properties. Section 5 presents the results of the model applied to the datasets presented in Section 2. Finally, Section 6 discusses the major contributions of this article and future work.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	The AHS is the largest survey conducted in Australia concerning health issues. The data used here is available through the mcglm package (Bonat 2018) in the ahs object. The main aim of this study is to investigate whether more access to health care services and demographic covariates such as sex, age, and income are related to the number of times patients use health services.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	The data has 5190 respondents and 15 variables; of which 10 were considered covariates and 5 as count response variables of a cross-sectional study with individuals over 18 years. The ﬁve response variables are Ndoc (Number of consultations with a doctor or specialist), Nndoc (Number of consultations with health professionals), Nadm (Number of admissions to a hospital, psychiatric hospital, nursing or convalescence home in the past 12 months), Nhosp (Number of nights in a hospital during the most recent admission) and Nmed (Total number of prescribed and non-prescribed medications used in the past two days). Table S1 in the supplementary material gives the description of each covariate.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	ship between the variables shows overdispersion for all variables, which Nhosp being the most overdispersed. This is characterized by Fisher Dispersion Index (DI) greater than 1 (Fisher 1934). Also, the generalized dispersion index (GDI) (Kokonendji and Puig 2018) classiﬁes this dataset as overdispersed once it is greater than 1 and its 95% conﬁdence interval does not contain zero. However, these results should be conﬁrmed by model ﬁtting.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	Figure 1: Barplot for the count of each response variable from Australian Health Survey (AHS) data. Ndoc (Number of consultations with a doctor or specialist), Nndoc (Number of consultations with health professionals), Nadm (Number of admissions to a hospital, psy- chiatric hospital, nursing or convalescence home in the past 12 months), Nhosp (Number of nights in a hospital during the most recent admission) and Nmed (Total number of prescribed and non-prescribed medications used in the past two days).
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	Table 1: Spearmean correlation, mean, variance, and dispersion index (DI) for the Aus- tralian Health Survey (AHS) response variables. The generalized dispersion index (GDI) and standard error (SE) equal 17.944 (1.99). Ndoc (Number of consultations with a doctor or specialist), Nndoc (Number of consultations with health professionals), Nadm (Number of admissions to a hospital, psychiatric hospital, nursing or convalescence home in the past 12 months), Nhosp (Number of nights in a hospital during the most recent admission) and Nmed (Total number of prescribed and non-prescribed medications used in the past two days).
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	Figure 2 presents the barplot for each response variable from ANT data. The 41 diﬀerent species present high variability. Some species were only seen one time in a single site, such as Polyrhachis and Solenopsis, while Iridomyrmex and Pheidole were seen 20 times on over one site.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	Figure 3 explores how the occurrence of diﬀerent species is related to each other by the Spearman correlation in a correlogram. The marginal correlation ranges from -.58 up to .74, which is well distributed along all potential values of the correlation parameter ρ = (−1, 1).
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	where each random eﬀect has mean 0, variance σ2, and correlation ρrr(cid:48)(r (cid:54)= r(cid:48)) between each pair of random eﬀects. This framework is general and can be applied to any distribution f and link function gr(.). Nevertheless, as we are dealing with count data, we considered only Poisson, binomial negative type II and COM-Poisson distributions, and a logarithm link function. Extending this model for diﬀerent distributions for each response (fr) is also possible, but we will not address it.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	Maximum likelihood is the estimation procedure, and it is fully described in da Silva et al. (2022). We integrated out the random eﬀects via numerical integration using Laplace ap- proximation (Tierney and Kadane 1986). The Newton-Raphson method was eﬃciently im- plemented to perform the inner optimization. We optimized the marginal likelihood using ﬁrst-order derivatives methods, such as BFGS and OPTIM routines available in the R software. The derivatives of the joint and marginal likelihood were obtained through automatic diﬀeren- tiation (Baydin, Pearlmutter, Radul, and Siskind 2017). This was eﬃciently implemented by the TMB package, which provides C++ templates where an objective function must be supplied, in this case, the log-likelihood function.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	The variability of the random variable in this model is being measured by two parameters, the variance of the random eﬀect and the dispersion of the pmf. A crucial point of the model is the great ﬂexibility to learn these two types of variances with no identiﬁability problems.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	The code used to produce the results of this paper is available on https://github.com/ guilhermeparreira/papers/tree/master/AJS_MGLMM_Overdispersed_Count and http:// www.leg.ufpr.br/doku.php/publications:papercompanions.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	In this section, we present simulation studies to assess the properties of the MLE estimators (bias and consistency). We considered a bivariate regression model for count data. We designed 12 simulation scenarios with four diﬀerent sample sizes, 100, 250, 500, and 1000, and three diﬀerent correlations between random eﬀects, ρ = −0.5, 0, 0.5. For the regression structure, we considered only an intercept for each response, with β01 = log(7) and β02 = log(1.5). The variance of random eﬀects were σ2 2 = .15. The dispersion parameter for NB and COM-Poisson was equal to φ = 1 and ν = .7 respectively, which induces a small overdispersion.
Processed_Multivariate_Generalized_Linear_Mixed_Models_for_C.txt	We generated 150, 200, and 300 datasets for Poisson, COM-Poisson, and NB distribution for each design. The primary idea was to generate 100 datasets for each distribution. However, as it did not return the SE of the parameter estimates in every repetition, it was necessary to increase the number of datasets generated proportionally to the number of SE failures for each distribution to obtain at least 100 valid estimations. The following three subsection presents the results for Poisson, NB, and COM-Poisson scenarios.

Subsets and Splits

No community queries yet

The top public SQL queries from the community will appear here once available.