The battalions were not confined to their locations and could be easily moved to locations far outside their home country. Since formation of the battalions was particularly slow in Belarus, many of them were first stationed there.  One of the first tasks of the battalions was mass execution of Jews. Attached to Einsatzgruppen as needed, the battalions rounded up, executed, and disposed of Jews. For example, it is estimated that Lithuanian Schutzmannschaft killed 78,000 Jews in Lithuania and Belarus.  The mass executions largely ceased by the end of 1941. By that time German advance into Soviet Union halted and Nazi officials considered using the battalions for more direct military duties. In particular, Franz Walter Stahlecker asked to relieve the 16th Army in the Demyansk Pocket .  However, Hitler refused. In Directive no. 46 , dated August 1942, he agreed to strengthen and enlarge Schutzmannschaft , but to use it only for anti-partisan operations and other auxiliary duties behind the front lines.  Some battalions continued to participate in the Holocaust (guarding or eliminating Jewish ghettos ).  The issue of involving Schutzmannschaft in combat was revisited after the Battle of Stalingrad . Some Schutzmannschaft battalions in Estonia, Latvia, Ukraine and elsewhere were reorganized into Waffen-SS divisions wearing national insignia.  Deserters were a constant problem for the battalions. For example, some 3,000 men deserted Lithuanian Schutzmannschaft between September 1943 and April 1944. 
An important topic in state-space modeling is model selection, or specifically to select the (discrete or continuous-valued) state dimensionality. Classical likelihood-based approaches rely on Akaike’s information criterion (AIC) or Bayesian information criterion (BIC), but these measures are often practically inefficient especially in the presence of sparse data samples. Following the Bayesian principle of “letting data speak for themselves”, the model selection problem has recently been tackled using nonparametric Bayesian inference, for instance in the cases of infinite HMM (Beal et al., 2002; Teh et al., 2006) and the switching SSM (Fox et al., 2010, 2011). Moreover, inference of a large-scale SSM for neuroscience data remains another important research topic. Exploiting the structure of the system, such as the sparsity, smoothness and convexity, may allow for employing efficient state-of-the-art optimization routines and imposing domain-dependent priors for regularization (Paninski et al., 2009; Paninski, 2010). Finally, developing consistent goodness-of-fit assessment for neuroscience data would help to validate and compare different statistical models (Brown et al., 2003).
Earwigs are generally nocturnal , and typically hide in small, dark, and often moist areas in the daytime. They can usually be seen on household walls and ceilings. Interaction with earwigs at this time results in a defensive free-fall to the ground followed by a scramble to a nearby cleft or crevice.  During the summer they can be found around damp areas such as near sinks and in bathrooms. Earwigs tend to gather in shady cracks or openings or anywhere that they can remain concealed during daylight. Some people erroneously believe that earwigs burrow into people's ears; that is mostly a myth, although earwigs may crawl into ears and some can bite, as other insects do. [ citation needed ] Picnic tables, compost and waste bins, patios, lawn furniture, window frames, or anything with minute spaces (even artichoke blossoms) can potentially harbour them.